Pajek. Program for Analysis and Visualization of Large Networks. Reference Manual List of commands with short explanation version 2.

Transcription

1 Pajek Program for Analysis and Visualization of Large Networks Reference Manual List of commands with short explanation version 2.05 Vladimir Batagelj and Andrej Mrvar Ljubljana, September 24, 2011

2 c 1996, 2011 V. Batagelj, A. Mrvar. Free for noncommercial use. PdfLaTex version October 1, 2003 Vladimir Batagelj Department of Mathematics, FMF University of Ljubljana, Slovenia Andrej Mrvar Faculty of Social Sciences University of Ljubljana, Slovenia

3 Contents 1 Pajek 3 2 Data objects 6 3 Main Window Tools File Net Nets Operations Partition Partitions Vector Vectors Permutation Permutations Cluster Hierarchy Options Info Tools Draw Window Tools Main Window Draw Tool Layout Layers GraphOnly Previous Redraw Next ZoomOut Options Export Spin Move Info Exports to EPS/SVG/X3D/VRML Defaults Parameters in EPS, SVG, X3D, and VRML Defaults Window

4 5.3 Exporting pictures to EPS/SVG defining parameters in input file Using Unicode in Pajek s pictures Using Macros in Pajek What is a Macro? How to record a Macro? How to execute the Macro? Example List of macros available in installation file Macros prepared for genealogies and other acyclic networks Macros prepared for computing derived kinship relations Repeating last command Blockmodeling in Pajek MDL files Examples of MDL files Regular blocks Diagonal blocks (clustering) Acyclic model (up) Acyclic model with symmetric clusters (down) Center-Periphery Regular path Regular chain mode standard model for Davis.net Colors in Pajek 89 9 Citing Pajek 91 2

5 Pajek Manual 3 1 Pajek Pajek is a program, for Windows, for analysis and visualization of large networks having some thousands or even millions of vertices. In Slovenian language the word pajek means spider. The latest version of Pajek is freely available, for noncommercial use, at its home page: We started the development of Pajek in November Pajek is implemented in Delphi (Pascal). Some procedures were contributed by Matjaž Zaveršnik. The main motivation for development of Pajek was the observation that there exist several sources of large networks that are already in machine-readable form. Pajek should provide tools for analysis and visualization of such networks: collaboration networks, organic molecule in chemistry, protein-receptor interaction networks, genealogies, Internet networks, citation networks, diffusion (AIDS, news, innovations) networks, data-mining (2-mode networks), etc. See also collection of large networks at: The design of Pajek is based on our previous experiences gained in development of graph data structure and algorithms libraries Graph and X-graph, collection of network analysis and visualization programs STRAN, RelCalc, Draw, Energ, and SGML-based graph description markup language NetML. Figure 1: Pajek/Spider

6 4 Pajek Manual global hierarchy local reduction cut-out context inter-links Figure 2: Approaches to deal with large networks The main goals in the design of Pajek are: to support abstraction by (recursive) decomposition of a large network into several smaller networks that can be treated further using more sophisticated methods; to provide the user with some powerful visualization tools; to implement a selection of efficient (subquadratic) algorithms for analysis of large networks. With Pajek we can: find clusters (components, neighbourhoods of important vertices, cores, etc.) in a network, extract vertices that belong to the same clusters and show them separately, possibly with the parts of the context (detailed local view), shrink vertices in clusters and show relations among clusters (global view). Besides ordinary (directed, undirected, mixed) networks Pajek supports also multi-relational networks, 2-mode networks (bipartite (valued) graphs networks between two disjoint sets of vertices), and temporal networks (dynamic graphs networks changing over time).

7 Pajek Manual 5 Figure 3: Pajek textbook This manual provides short explanations of all procedures implemented in the last version of Pajek. The novice users we advise to read the Pajek textbook [31] de Nooy W., Mrvar A., Batagelj V. (2002) Exploratory Social Network Analysis With Pajek. Structural Analysis in the Social Sciences 27, Cambridge University Press, For an overview of network analysis with Pajek see the NICTA workshop slides [5].

8 6 Pajek Manual 2 Data objects In Pajek six types of objects are used: Figure 4: Pajek s Main Window 1. Networks main objects (vertices and lines). Default extension:.net. Network can be presented on input file in different ways: using arcs/edges (e.g. 1 2 line from 1 to 2) using arcslists/edgeslists (e.g line from 1 to 2 and from 1 to 3) matrix format UCINET, GEDCOM, chemical formats... Additional information for network drawing can be included in input file as well. This is explained in the section Exports to EPS/SVG/VRML. 2. Partitions they tell for each vertex to which class vertex belong. Default extension:.clu. 3. Permutations reordering of vertices. Default extension:.per. 4. Clusters subset of vertices (e.g. one class from partition). Default extension:.cls. 5. Hierarchies hierarchically ordered vertices. Example: Root g1 g2 g11 g12 v5,v6,v7 v1,v2 v3,v4

9 Pajek Manual 7 Root has two subgroups g1 and g2. g2 is a leaf cluster with vertices v5,v6 and v7. g1 has two subgroups g11 and g12... Default extension:.hie. 6. Vectors they tell for each vertex some numerical property (real number). Default extension:.vec. By double clicking on selected network, partition,... you can show the object on screen. The procedures in Pajek s main window (see Figure 4) are organized according to the types of data objects they use as input. Permutations, partitions and vectors can be used to store properties of vertices measured in different scales: ordered, nominal (categorical) and numeric. Figure 5: Spider web; Photo: Vladimir Batagelj.

10 8 Pajek Manual 3 Main Window Tools 3.1 File Input/Output manipulation with the six data objects. Network N Read Read network from Ascii file. Edit Edit network. Choose vertex, show its neighbors and then: add new lines to/from selected vertex (by left mouse double clicking on Newline); delete lines (by left mouse double clicking); change value of line (by single right mouse clicking); subdivide line to two orthogonal lines using new invisible vertex (by single middle mouse clicking). Save Save selected network to Ascii file. If network represents Ore graph with the following five relations (arcs): 1. Wi Hu, 2. Mo Da, 3. Mo So, 4. Fa Da, 5. Fa So it can be stored as GEDCOM file. The other possibility is Pajek Ore graph: 1.Fa Ch, 2.Mo Ch, 3.Hu- Wi (edge), or 1.Pa Ch, 3.Hu-Wi (edge). Export Matrix to EPS write matrix in EPS format: Original using default numbering (for 1-mode and 2-mode networks). Using Permutation using current permutation. Additionally lines can be drawn to divide different classes defined by selected partition. Option can be used for 1-mode and 2-mode networks. Using Partition using current partition. In the text window number and density of lines among classes (and vertices in selected two classes) are displayed. Additionally matrix is exported to EPS where density is expressed using shadowing: 1. Structural Densities are normalized according to maximum possible number of lines among classes (suitable for dense networks). 2. Delta Densities are normalized according to vertices having the highest number of input and output neighbors in classes (suitable for sparse networks).

11 Pajek Manual 9 Diamonds for Negative Values, Circles for 0 Squares are used for posititive values, diamonds for negative and circles for value 0 (useful for black and white printing). Diamonds, Circles and Lines in GreyScale Diamonds, circles and dividing lines are drawn in greyscale (not in red, green and blue). Labels on Top/Right Labels are written on the top and on the right of the matrix - suitable for longer labels. Only Black Borders All squares in matrix have black borders, otherwise dark squares will have white and light squares will have black borders. Thick Boundary Line Use thicker line for dividing clusters. Large Squares/Diamonds/Circles Use larger or smaller squares, diamonds, and circles. Use Partition Colors for Vertex Labels Labels of vertices are displayed using partition colors. Change Label of selected network. Dispose selected network from memory. Time Events Network N Read Time Events Read time network described using time events. See Table 1. List of properties s can be empty as well. If several edges (arcs) can connect two vertices, additional tag like :k (k-th line) must be given to determine to which line the command applies. E.g. command HE: results in hiding the third edge connecting vertices 14 and 37. Example of time network described using time events: *Vertices 3 *Events TI 1 AV 2 "b" TE 3 HV 2 TI 4 AV 3 "e" TI 5 AV 1 "a" TI 6 AE TI 7 SV 2 AE 1 2 1

12 10 Pajek Manual Table 1: List of time events. Event TI t TE t AV vns HV v SV v DV v AA uvs HA uv SA uv DA uv AE uvs HE uv SE uv DE uv CV vs CA uvs CE uvs CT uv CD uv PE uvs AP uvs DP uv EP uvs Explanation initial events following events happen when time point t starts end events following events happen when time point t is finished add vertex v with label n and properties s hide vertex v show vertex v delete vertex v add arc (u,v) with properties s hide arc (u,v) show arc (u,v) delete arc (u,v) add edge (u:v) with properties s hide edge (u:v) show edge (u:v) delete edge (u:v) change vertex property change property of vertex v to s change arc property change property of arc (u,v) to s change edge property change property of edge (u:v) to s change type change (un)directedness of line (u,v) change direction of arc (u,v) replace pair of arcs (u,v) and (v,u) by single edge (u:v) with properties s add pair of arcs (u,v) and (v,u) with properties s delete pair of arcs (u,v) and (v,u) replace edge (u:v) by pair of arcs (u,v) and (v,u) with properties s

13 Pajek Manual 11 TE 7 DE 1 2 DV 2 TE 8 DE 1 3 TE 10 HV 1 TI 12 SV 1 TE 14 DV 1 See also other possibility: description of time network using time intervals. Save Save time network in time events format. Partition C Read partition from Ascii file. Edit partition (put vertices to classes). Save selected partition to Ascii file. Change Label of selected partition. Dispose selected partition from memory. Permutation P Read permutation from Ascii file. Edit permutation (interchange positions of two vertices). Save selected permutation to Ascii file. Change Label of selected permutation. Dispose selected permutation from memory. Cluster S (list of selected vertices) Read cluster from Ascii file. Edit cluster (add and delete vertices). Save selected cluster to Ascii file. Change Label of selected cluster. Dispose selected cluster from memory. Hierarchy H Read hierarchy from Ascii file.

14 12 Pajek Manual Edit hierarchy (change types and names of nodes, or show vertices (and subtree) belonging to selected node). Nodes can be pushed up and down within hierarcy. Save selected hierarchy to Ascii file. Change Label of selected hierarchy. Dispose selected hierarchy from memory. Vector V Read vector from Ascii file. Edit vector (change components of vector). Save selected vector(s) to Ascii file. If cluster representing vector id s is present, all vectors with corresponding id numbers will be saved to the same output file. Vector s id can be added to cluster by pressing V on the selected vector (empty cluster should be created first). All vectors must have the same dimensions. Change Label of selected vector. Dispose selected vector from memory. Pajek Project File *.paj Read Pajek project file (file containing all possible Pajek data objects networks, partitions, permutations, clusters, hierarchies and vectors). Save all currently loaded objects as a Pajek project file. Repeat session During program execution all commands are written to file *.log. In this way you can repeat any execution by running selected log file. If you change in the log file a name of a file to?, program will ask for name when running logfile next time (so you can repeat the same sequence of steps logfile with different input data). If startup logfile (Pajek.log) exists (in the same directory as Pajek.exe), it is automatically executed every time when Pajek is run. Show Report Window Bring the report window in the front in the case that it was closed or is not visible. Exit program.

15 Pajek Manual Net Operations, for which only a network is needed as input. Transform Transpose Transposed network of selected network: 1-Mode - Change direction of arrows. 2-Mode - Interchange Rows and Cols. Remove Selected Vertices Remove selected vertices from network. all Edges Remove all edges from selected network. all Arcs Remove all arcs from selected network. Multiple Lines Remove all multiple lines from selected network. 1. Sum Values Values of all deleted lines are added to not deleted line between corresponding two vertices. 2. Number of Lines Value of line between two vertices in a new network correspond to the number of lines between the two vertices in original network. 3. Min Value Minimum value of all lines between two vertices is selected. 4. Max Value Maximum value of all lines between two vertices is selected. 5. Single Line Value of line between two vertices in a new network is 1. Loops Remove all loops from selected network. Lines with Value 1. lower than Remove all lines with value lower than specified value. 2. higher than Remove all lines with value higher than specified value. 3. within interval Remove all lines with values within specified interval. all Arcs from each Vertex except 1. k with Lowest Line Values Sort lines around vertices in ascending order according to output line values. Keep only selected number of lines with lowest values.

16 14 Pajek Manual 2. k with Highest Line Values Sort lines around vertices in descending order according to output line values. Keep only selected number of lines with highest values. 3. keep Lines with Value equal to the k-th Value Determine what to do with lines with value equal to the k-th value (remove them or not). Triangle Remove arcs belonging to lower or upper triangle. Add additional vertices, lines or vertices/lines labels to network. Vertices Copy network to new network. Dimension can be enlarged for selected number of vertices (additional vertices without lines are added). Source and Sink If network is acyclic, add unique first and last vertex (new network has two artificial vertices). Default Vertex Labels Replace current labels of vertices by default vertices labels (v1, v2...). Vertex Labels from File Replace the default vertices labels (v1, v2...) by labels given in a file. Line Labels as Line Values replace labels of lines (or create new if there are no) with line values. Number of decimal places is the same as used in Draw window for marking lines with line values. Sibling edges Add sibling edges to vertices with a common 1. Input arc-ancestor 2. Output arc-descendant Edges Arcs Convert all edges to arcs (in both directions) (make directed network). Arcs Edges All Convert all arcs to edges (make undirected network). Bidirected only Convert only arcs in both directions to edges: 1. Sum Values Value of the new edge is the sum of values of both arcs. 2. Min Value Value of the new edge is the smaller of values of arcs. 3. Max Value Value of the new edge is the larger of values of arcs. Bidirected Arcs Arc

17 Pajek Manual 15 Select Min Value If there exist bidirected arcs between two vertices retain only the arc with lower value and remove the arc with higher value. If both values are equal replace both arcs with an edge. Select Max Value If there exist bidirected arcs between two vertices retain only the arc with higher value and remove the arc with lower value. If both values are equal replace both arcs with an edge. Line Values Transformations of line values: Recode Display frequency distribution of line values according to selected intervals and recode line values in this way. Multiply by a constant. Add Constant to line values. Constant min or max of line value and selected constant. Absolute line values. Absolute + Sqrt square root of line values. Truncate truncated line values. Exp exponent of line values. Ln natural logarithm of line values. Power selected power of line values. Normalize Reduction 1. Sum normalize so that the sum of line values will be 1 2. Max normalize so that the maximum line value will be 1 Degree (Recursively) delete from network all vertices with degree lower than selected value (according to Input, Output or All degree). Operation can be limited to selected cluster. Hierarchical Recursively delete from network all vertices that have only 0 or 1 neighbor. Results: simpler network and hierarchy with deleted vertices. Original network can be later restored (if we forget directions of lines). Subdivisions Recursively delete from network all vertices that have exactly 2 neighbors (together with corresponding two lines) and (instead of that) add direct line between these two neighbors. Result is simpler network (for drawing). Original network cannot be restored!

18 16 Pajek Manual Figure 6: Part of Reuters Terror News network on the 36th day. Design (flow graph) Reduction of all structural parts of network according to McCabe (for programs flow graphs) [50]. Generate in Time Generate network in specified time(s) or interval. Input first time, last time and step (integers). Additional parameters when vertices and lines are active should be given in network to perform this operation. They must be given between signs [ and ]: - is used to divide lower and upper limit of interval,, is used to separate intervals, * means infinity. Example: *Vertices 3 1 "a" [5-10,12-14] 2 "b" [1-3,7] 3 "e" [4-*] *Edges [7]

19 Pajek Manual [6-8] Vertex a is active from times 5 to 10, and 12 to 14, vertex b in times 1 to 3 and in time 7, vertex e from time 4 on. Line from 1 to 2 is active only in time 7, line from 1 to 3 in times 6 to 8. The lines and vertices in a temporal network should satisfy the consistency condition: if a line is active in time t then also its end-vertices are active in time t. When generating time slices of a given temporal network only consistent lines are generated. Note that time records should always be written as last in the row where vertices / lines are defined. See also other possibility of describing time network: description of time network using time events. All Generate all networks in specified times. Only Different Generate network in specified time only if the new network will differ in at least one vertex or line from the last network which was generated. Interval Generate network with vertices and lines present in selected interval. 1-Mode to 2-Mode Generate 2-mode network from any network. 2-Mode to 1-Mode Generate an ordinary (1-mode) network from 2-mode (affiliation) network. Result is a valued network. To store a 2-mode network in input file use Pajek or Ucinet format (look at Davis.dat from Ucinet dataset). Rows Result is a network with relations among row elements (actors). The value of line tells number of common events of the two actors. Columns Result is network with relations among column elements (events). The value of a line tells number of actors that took part in both events. Include Loops If checked, loops with value telling the total number of events for each actor (total number of actors for each event), are added. Multiple Lines Generate nonvalued 1-mode network, where multiple lines among vertices can exist. The label of the generated line corresponds to the label of the event/actor that served to induce the line. If partition of the same dimension is present, multirelational network can be generated.

20 18 Pajek Manual Normalize 1-Mode Normalize the obtained 1-Mode network. 1-Mode network must be obtained with option include loops checked, and multiple lines not checked: Geo ij = a ij aii a jj Input ij = a ij a jj Output ij = a ij a ii Min ij = Max ij = MinDir ij = MaxDir ij = a ij min(a ii, a jj ) a ij max(a ii, a jj ) { aij a ii a ii a jj 0 otherwise { aij a jj a ii a jj 0 otherwise The obtained network is usually not sparse. To make it sparser use Net/Transform/Remove/lines with value/lower than. Rows=Cols Transform 2-Mode network with the same subsets of vertices to 1-Mode network. Cols=0 Transform 2-Mode network to 1-Mode network by setting number of columns to 0. The result is the same as changing for example *Vertices to *Vertices 32 in input network file. Multiple Relations Extract Relation(s) Extract one or selected list of relations from selected multiple relations network. Canonical Numbering Enumerate relations with sequential numbers 1, 2,... Generate 3-Mode Network generate a 3-mode network from 1-mode or 2-mode multirelational network. For each line in multirelational network r: i j v (line from i to j with value v, relation number is r) generate the following three lines (triangle): 1-mode networks: i N+j v

21 Pajek Manual 19 i 2N+r v N+j 2N+r v 2-mode networks: i j v i N+M+r v j N+M+r v where N is cardinality of the first mode and M cardinality of the second mode. Line Values > Relation Numbers Store line values as relation numbers (absolute truncated values). Relation Numbers > Line Values Store relation numbers as line values. Change Relation Number - Label Change selected relation number to new relation number with corresponding label. Sort Lines Neighbors around Vertices For each vertex sort lines connected to it in ascending order according to other end-vertex. Line Values Sort lines in ascending or descending order according to line values. Random Network Generate random network of selected dimension Total No. of Arcs Generate random directed network of selected dimension and given number of arcs. Vertices Output Degree Generate random directed network of selected dimension and output degree of each vertex in given range. Bernoulli/Poisson Generate undirected, directed, acyclic, bipartite or 2-mode random network according to model defined by Bernoulli / Poisson: each line is selected with the given probability p. Instead of p, which is for large and sparse networks (very) small number, in Pajek a more intuitive average degree d is used. They are connected with relations d = 1 n v V deg(v) = 2m and m = pm where n = V, n m = L and M is the number of lines in maximal possible network for example, for undirected graphs M = n(n 1). Scale Free Generate scale free undirected, directed or acyclic network. The procedure is based on a refinement of the model for generating scale free networks, proposed in [55]. At each step of the growth a new vertex and k edges are added to the network N. The endpoints

22 20 Pajek Manual of the edges are randomly selected among all vertices according to the probability Pr(v) = α indeg(v) E + β outdeg(v) E + γ 1 V where α + β + γ = 1. It is easy to check that v V Pr(v) = 1. Small World Generate Small world random network. See [12]. Extended Model Generate random network according to extended model defined by Albert and Barabasi [3]. Partitions Partitioning Network. Result is a Partition. Degree Input Number of lines into vertices. Output Number of lines out of vertices. All Number of neighbors of vertices. Domain For each vertex compute its domain according to input, output or all neighbors. Results are: Partition containing size of domain - number of reachable vertices. Vector containing the normalized size of domain - normalization is done by total number of vertices 1. Vector containing the average distance from/to domain. Proximity Prestige index can be computed by dividing the normalized size of domain by average distance. Core k-core is a subnetwork of given network where each vertex has at least k neighbors in the same core according to: Input... lines coming into vertex. Output... lines going out of vertex. All... all neighbors. 2-Mode core partition of a 2-mode network. Given minimum degree in first (k 1 ) and minimum degree in second subset (k 2 ) a new partition is generated where 0 means that vertex does not belong to the core of prespecified k 1 and k 2, 1 means that vertex belongs to that core.

23 Pajek Manual Figure 7: US Patents Main island liquid-crystal display

24 22 Pajek Manual 2-Mode Review Given starting values of k 1 and k 2 the following list is computed: k 1 k 2 Rows Cols Comp where k 1 is minimum degree in the first, k 2 minimum degree in the second subset, Rows and Cols are number of vertices in first and second subset respectivelly and Comp, number of connected components in network induced by k 1 and k 2. k 1 and k 2 are incremented until the resulting network is empty. 2-Mode Border Compute only border values of k1 and k2 for a given 2-mode network. Valued Core Generalized k-core: Instead of counting lines (neighbors) use values of lines. sum of lines or maximum value can be used when computing valued core: Sum valued core of threshold val is a subnetwork of given network where the sum of values of lines to (from) the members of the same core is at least val. Max valued core of threshold val is a subnetwork of given network where the maximum value of all lines to (from) the members of the same core is at least val. Threshold values must be given in advance. Two different ways to determine thresholds: First Threshold and Step Select first threshold value and step in which to increase threshold. Selected Thresholds Thresholds (increasing numbers) are given using vector. 2-Mode valued core (according to line values) partition of a 2- mode network. Given minimum valued degree in first (k 1 ) and minimum valued degree in second subset (k 2 ) a new partition is generated where 0 means that vertex does not belong to the valued core of prespecified k 1 and k 2, 1 means that vertex belongs to that core. Additionally (for 1-mode networks), Input, Output or All valued cores can be used. Depth Acyclic Partition acyclic network according to depths of vertices. Genealogical Partition network that represents genealogy according to layers of vertices.

25 Pajek Manual 23 Generational Partition network that represents genealogy according to layers of vertices. The same as genealogical partition but with less layers. p-cliques Partition network according to p-cliques (partition to clusters where vertices have at least proportion p (number between 0 and 1) neighbors inside the cluster. Strong... for directed network. Weak... for undirected network. Vertex Labels Partition vertices with same labels to the same class numbers (for molecule). Vertex Shapes Partition vertices with same shapes (ellipse, box, diamond) to the same class numbers (used in genealogy to show gender). Islands Partition vertices of network with values on lines (weights) to cohesive clusters (weights inside clusters must be larger than weights to neighborhood): the height of vertex (vector) is defined as the maximum weight of the neighbor lines. Two options: Line Weights Line Weights [Simple] New network with only lines constituting islands can be generated if Generate Network with Islands is checked. Bow-Tie Partition vertices of directed network (graph structure of the web) to the following classes: 1 LSCC, 2 IN, 3 OUT, 4 TUBES, 5 TENDRILS, 0 OTHERS. 2-Mode Partition of vertices of a 2-mode network into two subsets. Default Labels Partition Input is network with default vertex labels: e.g., v3, v9,... Result is a partition of selected dimension, where vertices defined by numbers stored in vertex labels (e.g., 3, 9,...) go to cluster 1, other vertices go to cluster 0. Operation can be used to make other objects (e.g. partitions, vectors,...) compatible with a network, if network is reduced by several operations (e.g. extractions). Components Strong Strong Components of selected network. Strong-Periodic Strong Periodic Components of selected network - strongly connected components are further divided according to periods.

26 24 Pajek Manual Figure 8: Bow-tie Graph structure in the web [26] Weak Weak Components of selected network. Bi-Components Biconnected Components of selected network. Articulation points belong to several classes, so the result cannot be stored in partition biconnected components are stored in hierarchy! Minimal number of vertices in components can be selected. Additionally, partition containing articulation points is produced: number of biconnected components to which each vertex belongs is given. Partition containing vertices belonging to exactly one bicomponent, vertices outside bicomponents and articulation points is also produced: vertices outside bicomponents get class zero, each bicomponent is numbered consecutively (from 1 to number of bicomponents) and articulation points get class number Hierarchical Decomposition Clustering* Hierarchical clustering procedure. Input is dissimilarity network (matrix), which can be obtained using Operations/Dissimilarity/Network based or read from input file. Run Hierarchical clustering procedure. Result is hierarchy with nested clusters and dendrogram in EPS.

27 Pajek Manual 25 Options Select method for hierarchical clustering procedure (general, minimum, maximum, average, ward, squared ward). Symmetric-Acyclic Symmetric-Acyclic decomposition of network. Result is hierarchy with nested clusters [33]. Clustering with Relational Constraint Hierarchical clustering with relational constraint procedure. See: Ferligoj A., Batagelj V. (1983): Some types of clustering with relational constraints. Psychometrika, 48(4), Only dissimilarities among vertices that are linked are taken into account what enables to find clusterings very fast also for large networks. Input is network with dissimilarities, which can be obtained using Operations/Dissimilarity/Network or Vector based or read from input file. Numbering Run Results are: a partition representing tree: fathers of nodes; and two vectors: describing heights of nodes and number of vertices in subtree respectivelly. If network has n vertices then obtained partitions and vectors have dimension 2*n-1. Note that this objects are not compatible with original network, you must use Make Partition to get compatible results. Make Partition From obtained partition representing tree generate partition compatible with original network using Threshold determined by Vector From obtained partition representing tree and one of the two vectors (all have dimension 2*n-1) generate partition compatible with original network by giving threshold value. with selected Size of Clusters From obtained partition representing tree and given number of vertices in clusters generate partition compatible with original network. Extract Subtree as Hierarchy Extract subtree from obtained Partition by giving the root as Pajek Hierarchy. Options Select method for hierarchical clustering with relational constraint (minimum, maximum, or average) and strategy (strict, leader, or tolerant). Depth First Depth first numbering of selected network... Strong... taking directions of arcs into account. Weak... forget directions (or undirected network).

28 26 Pajek Manual Breadth First Breadth first numbering of selected network... Strong... taking directions of arcs into account. Weak... forget directions (or undirected network). Reverse Cuthill-McKee RCM numbering. See paper. Core + Degree Numbering in decreasing order according to all core partition. Within the same core number vertices are ordered in decreasing order according to number of neighbors which have the same or higher core number. Citation Weights If a network represents citation network, weights of lines (citations) and vertices (articles) can be computed. Results are: Network with values on lines representing importance of citations. Binary partition with vertices on the main path. Network containing only main path. Vector with importance of vertices (articles). Different methods of assigning weights [43]: Search Path Count (SPC) method. Compute from Source to Sink. Search Path Link Count (SPLC) method. Each vertex is considered as Source. Search Path Node Pair (SPNP) method. Weights can also be normalized (using flow or maximum value) or logged. k-neighbors Select all vertices Input...from which we can reach selected vertex in at most k-steps. Output...that can be reached from selected vertex in at most k-steps. All...Input + Output (forget direction of lines) Result is partition where vertices are in class numbers equal to the distance from given vertex, vertices that cannot be reached from selected vertex are in class number After you have a partition you can extract subnetwork. From Clusters Compute selected distances according to each vertex in Cluster. Results consist of so many partitions as is the number of vertices in cluster. Instead of storing results in partitions they can be stored in vectors as well.

29 Pajek Manual 27 Paths between 2 vertices One Shortest Find the shortest path between two vertices. Result is new network. Values on lines can be taken into account (if they present distances between vertices) or not (graph theoretical distance). The latter possibility is faster. All Shortest Find all shortest paths between two vertices. Result is new network. Values on lines can be taken into account (if they present distances between vertices) or not (graph theoretical distance). The latter possibility is faster. Walks with Limited Length Find all walks between two vertices with limited maximum length. Diameter Find diameter the length of the longest shortest path in network and corresponding two vertices. Full search is performed, so the operation may be slow for very large networks (number of vertices larger than 2000). Geodesics Matrices* Compute the shortest path length matrix and the geodesics count matrix (for small networks only!). Distribution of Distances Compute distribution of lengths of the shortest paths and average path length among all reachable pairs of vertices in network. From All Vertices Take all vertices as starting points. From Vertices in Cluster Only distances from vertices selected by Cluster are computed. Critical Path Method (CPM) Find the critical path in acyclic network result is new network containing the critical path. Algorithm can be used in the area of project planning but also for analysing acyclic graphs. Additional networks containing total and free delay times of activities are generated. Two vectors (partitions) are generated, too: First containing the earliest possible times of coming into given states and the second containing the latest feasible times of coming into given states. Maximum Flow among vertices. Selected Pair Find maximum flow between selected two vertices (algorithm looks for paths to be saturated and among them it always selects the shortest path). Algorithm can be used in the technical area (actual flow, values on lines mean capacities) or for analysing graphs (if all values are 1). Result is a new network, containing the two vertices and lines contributing to maximum flow between them.

30 28 Pajek Manual Pairs in Cluster Find maximum flow among vertices determined by cluster. Result is a new network, where a value on line means maximum flow between corresponding two vertices. Algorithm is slow: Use it on smaller networks or clusters with limited number of vertices only! Vector Get vector from network Centrality Result is a vector containing selected centrality measure of each vertex and centralisation index of the whole network [64, p ]. Closeness centrality (Sabidussi). 1. Input centrality of each vertex according to distances of other vertices to selected vertex. 2. Output centrality of each vertex according to distances of selected vertex to all other vertices. 3. All forget direction of lines consider network as undirected. Betweenness centrality (Freeman). Get Loops store values of loops to vector. Get Coordinate x, y, or z coordinate of network. You can also get all coordinates at once - possibility to have more than 3 coordinates, coordinates must contain character. (dot). Important Vertices Find important vertices in directed network (e.g. web pages, scientific citations) or 2-mode network. Result are vectors with weights and partition with selected number of important vertices. 1-Mode: Hubs-Authorities In directed networks we can usually identify two types of important vertices: hubs and authorities [47]. A vertex is a good hub, if it points to many good authorities, and it is a good authority, if it is pointed to by many good hubs. In obtained partition value 1 means, that the vertex is a good authority, value 2 means, that the vertex is a good authority and a good hub, and value 3 means, that the vertex is a good hub. 2-Mode: Important Vertices Generalization of algorithm for 2-mode networks find important vertices from first and second subset. Structural Holes Burt s measure of constraint (structural holes) [27, page 54-55]. Results are:

31 Pajek Manual 29 network p ij : the proportion of the value of i s relation(s) with j compared to the total value of all relations of i. where a ij is the value of the line from i to j p ij = a ij + a ji k (a ik + a ki ) network containing dyadic constraint c ij the constraint of absent primary holes around j on i: Explanation: Contact j constrains your i s entrepreneurial opportunities to the extent that: (a) you ve made a large investment of time and energy to reach j, and (b) j is surrounded by few structural holes with which you could negotiate to get a favorable return on the investment. c ij = (p ij + k,k i,k j p ik p kj ) 2 vector containing aggregate constraint C i : C i = j c ij, C i = 1 for isolated vertices. Clustering Coefficients Compute different inherent tendency coefficients in undirected network: Let deg(v) denotes degree of vertex v, E(G 1 (v)) number of lines among vertices in 1-neighborhood of vertex v, MaxDeg maximum degree of vertex in a network, and E(G 2 (v)), number of lines among vertices in 1 and 2-neighborhood of vertex v. CC 1 coefficients considering only 1-neighborhood: CC 1 (v) = 2 E(G 1 (v)) deg(v) (deg(v) 1) CC 1(v) = deg(v) MaxDeg CC 1(v) CC 2 coefficients considering 2-neighborhood CC 2 (v) = E(G 1(v)) E(G 2 (v)) CC 2(v) = deg(v) MaxDeg CC 2(v) If deg(v) 1 all coefficients for vertex v get missing value ( ). Watts-Strogatz Clustering Coefficient (Transitivity) and Network Clustering Coefficient are also reported. Summing up Values of Lines Sum values of all incoming, outgoing or all lines connected to selected vertex.

32 30 Pajek Manual Min of Values of Lines Find minimum value of incoming, outgoing or all lines connected to selected vertex. Max of Values of Lines Find maximum value of incoming, outgoing or all lines connected to selected vertex. Centers Find centers in a graph using robbery algorithm: vertices that have higher degrees (are stronger) than their neighbors steal from them: at the beginning give to vertices initial strength according to their degrees, or start with value 1 when weak vertex is found, neighbors steal from it according to their strengths, or they steal the same amount PCore generalized cores. Degree ordinary cores. Sum taking values of lines into account (sum of values of lines inside pcore). Max taking values of lines into account (max of values of lines inside pcore). Count - how many times each line belongs to predefined rings 3-Rings For each line count number of 3-rings to which the line belongs. Undirected for undirected networks count undirected 3-rings. Directed for directed networks count cyclic, transitive, or all 3-rings, or count how many times each line is a transitive shortcut (see Figure 9). 4-Rings For each line count number of 4-rings to which the line belongs. Undirected for undirected networks count undirected 4-rings. Directed for directed networks count cyclic, diamonds, genealogical, transitive, or all 4-rings, or count how many times each line is a transitive shortcut (see Figure 10). 3.3 Nets Operations on two networks.

33 Pajek Manual 31 cyclic transitive Figure 9: Lines belonging to cyclic and transitive (shortcut) 3-rings cyclic transitive genealogical diamond Figure 10: Types of directed 4-rings on arcs Union of lines Fuse selected networks. Result is a multiple relations network. If you want to get union of networks, multiple lines must still be deleted. Networks must match in dimension or: If one network has m vertices and other n vertices and m < n then in network with n vertices first m vertices must match with vertices in network with m vertices. Cross-Intersection Intersection of selected networks. Networks must match in dimension or: If one network has m vertices and other n vertices and m < n then in network with n vertices first m vertices must match with vertices in network with m vertices. Values of lines in intercept can be sum, difference, product, quotient, min, or max of both values. Intersection Intersection of selected networks where relation numbers are taken into account. Cross-Difference Difference of selected networks. Difference Difference of selected networks where relation numbers are taken into account. Union of vertices Add the second network at the end of first network.

34 32 Pajek Manual Fragment (1 in 2) Find all instances of fragment (determined by network 1) in network 2. Find Execute command. Options Select appropriate model of fragment. Induced there should be no additional lines between vertices in instance of fragment to match (stronger condition) otherwise additional lines can be present (weaker). Labeled labels must match (e.g. atoms in molecule). Labels are determined by classes (colors) in partition - first partition and second partition must be selected before searching for labeled fragments. First partition determines labels of first network (fragment), second partition determines labels of second (original) network. Check values of lines values of lines must match (e.g. in genealogy values represent sex: 1 man, 2 woman). Check relation number relation numbers must match. Check only cluster only fragments are searched. where first vertex is one of the vertices in cluster. Extract subnetwork produce additional result: extract subnetwork containing vertices belonging to fragments and corresponding lines. Retain all vertices after extracting in extracted network the same vertices as in original network are present, only lines which do not belong to any fragment are removed. Same vertices determine one fragment at most how fragments on the same set of vertices are treated: if not checked fragments with the same set of vertices are allowed. Create Hierarchy with fragments result of fragment searching is also a Hierarchy with vertices in fragments (available only if Same vertices determine one fragment at most is not checked. Repeating vertices in fragment allowed same vertices can appear in fragment more than once (e.g. in cycles). if not checked: found fragments always have the same number of vertices as original fragment if checked: some of found fragments can have less vertices than original fragment

35 Pajek Manual 33 Michael/Zrieva/ Francischa/Georgio/ Junius/Georgio/ Anucla/Zrieva/ Nicola/Ragnina/ Nicoleta/Zrieva/ Marinus/Zrieva/ Maria/Ragnina/ Damianus/Georgio/ Legnussa/Babalio/ Junius/Zrieva/ Margarita/Bona/ Lorenzo/Ragnina/ Slavussa/Mence/ Nicolinus/Gondola/ Franussa/Bona/ Sarachin/Bona/ Nicoletta/Gondola/ Marin/Gondola/ Magdalena/Grede/ Marinus/Bona/ Phylippa/Mence/ Figure 11: Fragments Marriages among relatives in Ragusa Multiply First * Second - multiply selected 1 or 2 mode networks (that match criteria for multiplication). Shrink coordinates (1 to 2) - Useful if you shrink network, draw shrunk network separately, and then apply all coordinates to vertices in original network (vertices in same class get the same coordinates). Replace coordinates in network 2 using coordinates of shrunk network 1. Shrinking can be determined using Partition or Hierarchy 3.4 Operations One network and something else is needed as input. Shrink Network - Before starting shrinking, select appropriate blockmodel in Options menu. Default is just number of lines between shrunk vertices that must be present in original network, to cause a line in a new network. Partition Shrink network according to selected partition. Vertices in class 0 are (by default) left unchanged, others are shrunk. Results are shrunken network and shrunken partition.

36 34 Pajek Manual Hierarchy Shrink network according to selected hierarchy. Nodes in hierarchy that are Closed are shrunk to new vertex. Cut nodes are shrunk to virtual vertex. Border nodes are not shrunk, but they are not visible. Vertices belonging to other nodes are left unchanged. Type of shrinking (blockmodel) can be selected in Options menu. Extract from Network Partition Extract sub-network according to selected partition (extract range of classes from partition). Extracted partition is produced as additional result. Cluster Extract sub-network according to selected cluster. 2-Mode Network Extract 2-mode network from 1-mode network: first and second mode are determined by given set of clusters in partition. to GEDCOM Extract sub-genealogy according to selected partition (weakly connected component) to new GEDCOM file (genealogy must be read as Ore graph). Brokerage Roles - For each vertex j count five brokerage roles (coordinator, itinerant broker, representative, gatekeeper and liaison) according to given partition. j j j j j i k i k i k i k i k coordinator itinerant broker representative gatekeeper liaison Dissimilarity* Network based Compute selected dissimilarity matrix (d 1, d 2, d 3 or d 4 ) among vertices in cluster according to number of common neighbors. Corrected Euclidean-like d 5 and Manhattan-like d 6 dissimilarities can be computed as well [13]. The obtained matrix can be used further for hierarchical clustering procedure. You can include vertex v to its own neighborhood or not and display in report window only upper triangle / undirected or complete matrix /directed (if number of vertices is low).

37 Pajek Manual 35 N v is a set of input, output or all neighbors of vertex v; + stands for symmetric sum, stands for set union and \ stands for set difference; stands for set cardinality; 1st maxdegree and 2nd maxdegree are the largest degree and the second largest degree in network, respectively. N u + N v d 1 (u, v) = 1st maxdegree + 2nd maxdegree d 2 (u, v) = N u + N v N u N v d 3 (u, v) = N u + N v N u + N v d 4 (u, v) = max( N u \ N v, N v \ N u ) max( N u, N v ) n d 5 (u, v) = d 6 (u, v) = s=1 s u,v ((q us q vs ) 2 + (q su q sv ) 2 ) + p ((q uu q vv ) 2 + (q uv q vu ) 2 ) n ( q us q vs + q su q sv ) + p ( q uu q vv + q uv q vu ) s=1 s u,v Dissimilarities d 5 and d 6 are based on some matrix Q = [q uv ] on vertices for example on adjacency matrix or on distance matrix. The parameter p is usually set to value 1 or 2. In the case N u = N v = 0 we set all dissimilarities d 1 - d 4 to 1. If Among all linked Vertices only is checked dissimilarities are computed as line values of given network. Vector based Euclidean, Manhattan, Canberra, or (1-Cosine)/2 dissimilarities among Vectors determined by Cluster are computed as line values of given network. Vector Operations on network and vector. Network * Vector Ordinary multiplication of matrix (network) by vector. Result is a new vector. Vector # Network Result is a new network: Input Multiplying incoming arcs in network by corresponding vector values - multiplying i-th column of matrix by i-th component of vector.

38 36 Pajek Manual Output Multiplying outgoing arcs in network by corresponding vector values - multiplying i-th row of matrix by i-th component of vector. Harmonic Function See Bollobas [25, page 328]. Let (G, a) be a connected weighted graph, with weight function a(x, y), and let S is subset of vertices V (G). A function f : V (G) IR is said to be harmonic on (G, a), with boundary S, if f(x) = 1 A(x) (a(x, y)f(y)), y x V (G) \ S A(x) = y a(x, y) Implementation in Pajek: function f is determined by vector weight function a(x, y) is given by (valued) network subset S is determined by partition vertices in class 1 are in subset S (fixed vertices), other vertices are in V (G) \ S additionally, permutation determines the order of vertices in computations. In Pajek you can compute the harmonic function once or iterativelly - as long as difference between successive functions become small enough. Components of vector that represents function f can be modified immediately when they are computed or only at the end of each iteration (after all components are computed). Procedure can be run according to: Input neighbors Output neighbors All neighbors Summing up neighbors For each vertex compute the sum of class numbers of its neighbors according to Input neighbors Output neighbors All neighbors Min of neighbors For each vertex compute the minimum class number of its neighbors according to Input neighbors

39 Pajek Manual 37 Output neighbors All neighbors Max of neighbors For each vertex compute the maximum class number of its neighbors according to Input neighbors Output neighbors All neighbors Put Loops put vector values as loops (arcs or edges) in current network. Put Coordinate put vector as x, y, or z coordinate, or put it as polar radius or polar angle of vertices in network layout. Diffusion Partition Compute diffusion partition according to thresholds given in vector. Vertices in selected cluster are considered to adopt in time 1. Islands Partition vertices to cohesive clusters according to weights of vertices determined by a vector. Vertex Weights Vertex island is a cluster of vertices of given network with weighted vertices where the weights of the vertices on the island are larger than the weights of the vertices in the neighborhood. The weights are also called heights. Vertex Weights [Simple] Simple vertex island is vertex island with only one top. Transform Transformations of network according to Partition, Cluster and/or Vector. Remove Lines Removing lines according to partition. Inside Clusters Remove all lines with incident vertices in the same (selected) cluster(s). Between Clusters Remove all lines with incident vertices in different clusters. Between Two Clusters 1. Arcs Remove all arcs pointing from first to second cluster. 2. Edges Remove all edges between the selected two clusters. Inside Clusters with value 1. lower than Vector value Remove all lines inside clusters (determined by a Partition) with value lower than the value specified in a Vector.

40 38 Pajek Manual 2. higher than Vector value Remove all lines inside clusters (determined by a Partition) with value higher than the value specified in a Vector. Dimension of a Vector must be equal to the highest cluster number in a Partition. Add some elements to network Arcs from Vertex to Cluster add arcs from selected vertex to all vertices in Cluster. Arcs from Cluster to Vertex add arcs from all vertices in Cluster to selected vertex. Time Intervals determined by Partitions change network to temporal network using two partitions: first partition determines initial time point, second determines terminal time point of each vertex. Direction Convert to directed network where all arcs are pointing from Lower->Higher class number. Higher->Lower class number. Lines inside classes may be deleted or not. Vector(s) -> Line Values Replace line values with result of selected operation (sum, difference, multiplication, division) on vector(s) values in corresponding terminal and initial vertices. Reorder Network Reorder vertices in network according to selected permutation. Partition Reorder vertices in partition according to selected permutation. Vector Reorder vertices in vector according to selected permutation. Count neighbor Colors For selected network and partition a new partition is generated where for each vertex the frequency of vertices of selected color in the neighborhood is given. Colors to be counted are determined using cluster. Coloring Create New Sequential coloring of vertices in order determined by permutation. Result depends on selected permutation significantly.

41 Pajek Manual 39 Complete Old Complete partial coloring of vertices in order determined by permutation. For example some vertices can be colored by hand, but most of the vertices are still uncolored (in class 0). In this way you can help program to produce better coloring. Balance* Relocation algorithm for partitioning signed graphs (graphs with positive and negative values on lines representing friends and enemies, for example). Given partition is optimized to get as much as possible positive lines inside classes and negative lines between classes. Another algorithm does not distinguish between diagonal and off-diagonal blocks: each block can be positive, negative, or null. If number of repetitions is higher than 1, initial partitions into given number of classes are chosen randomly for every repetition separately. If program finds several optimal solutions, all are reported. For more details about algorithm see Doreian and Mrvar [32]. Option can be used for two mode signed graphs as well: input is two mode partition. In this case algorithm tries to find as clear as possible positive, negative, and null blocks. If Prespecification is checked user can define a prespecified model by entering letters P, N, or 0 to cells (to require positive, negative or null blocks) or leave cells empty (in this case the block can be of any type). By setting penalty for small null blocks to some nonzero value, we try to get null blocks as large as possible. Blockmodeling* Generalized blockmodeling of 1-mode and 2-mode networks [7, 35]. For details see Section 7 on page 84. Descriptions of models are stored on MDL files. See also block types on page 50. Random Start Start the optimization with random partition(s). Optimize Partition Show the criterion function for selected partition and optimize it. Restricted Options Show only selected part of options (sufficient for most users) or all options. Short Report Show only main results of optimization in Report window (sufficient for most users) or detailed, long report. Genetic Structure Compute genetic structure of given acyclic network according to given partition (of minimal vertices). As result we get as many vectors as is different clusters in partition, and the dominant gene partition. Permutation* Improve given permutation according to network.

42 afg alb alg arg aus aut bel bol bra brm bul bur cam can car cha chd chi col con cos cub cyp cze dah den dom ecu ege egy els eth fin fra gab gha gre gua gui hai hon hun ice ind ins ire irn irq isr ita ivo jam jap jor ken kmr kod kor kuw lao leb lib liy lux maa mat mex mla mli mon mor nau nep net nic nig nir nor nze pak pan par per phi pol por rum rwa saf sau sen sie som spa sri sud swe swi syr tai tha tog tri tun tur uga uki upv uru usa usr ven vnd vnr wge yem yug zai afg alb alg arg aus aut bel bol bra bul bur can car cha chd chi col con cos cub cyp cze dah den ecu ege egy els eth fin fra gab gha gre gua gui hai hon hun ice ind ins ire irn irq isr ita ivo jam jap jor ken kmr kod kor lao leb lib liy lux mat mla mli nau nep net nic nig nir nor nze pak pan par per phi pol por rwa saf sau sen sie spa sri sud 40 Pajek Manual Pajek - shadow 0.00,1.00 Sep World trade - alphabetic order brm swi syr tai tha tog tri tun tur uga uki upv uru usa usr ven vnd vnr yug zai uki net bel lux fra ita den jap usa can bra arg ire swi spa por wge ege pol aus hun cze yug gre bul rum usr fin swe nor cam irn tur irq egy dom leb cha ind pak kuw maa aut cub mex uru mex nig ken saf mor mon mor sud syr isr sau rum kuw sri som tha mla swe wge yem gua hon els nic cos pan col ven ecu per chi tai kor vnr phi ins nze mli sen nir ivo upv gha cam gab maa alg hai dom jam tri bol par mat alb cyp ice dah nau gui lib sie tog car chd con zai uga bur rwa som eth tun liy jor yem afg mon kod brm nep kmr lao vnd uki net bel lux fra ita den jap usa can bra arg ire swi spa por ege pol aus hun cze yug gre bul usr fin nor irn tur irq egy leb cha ind pak aut cub uru nig ken saf sud syr isr sau Pajek - shadow 0.00,1.00 Sep World Trade (Snyder and Kick, 1979) - cores wge rum swe mex mor sri tha mla gua hon els nic cos pan col ven ecu per chi tai kor vnr phi ins nze mli sen nir ivo upv gha kuw gab cam alg hai maa jam tri bol par mat alb cyp ice dah nau gui lib sie tog car chd con zai uga bur rwa dom eth tun liy jor som afg kod yem nep kmr lao vnd mon brm Figure 12: World trade. Orderings: alphabetical and determined by clustering Travelling Salesman Can be applied to dissimilarity matrix, or modified matrix representing network (fill diagonal and change 0 in the matrix with some large numbers): Run Run 3-OPT algorithm for solving Travelling Salesman Problem. Options Put selected value on diagonal, add some artificial vertices, and incident lines with large values, change value 0 with selected (large) value. Seriaton Starting with network and (random) permutation improve the permutation using seriation algorithm from Murtagh [53, page 11-16]. 1-Mode for ordinary (1-Mode) networks 2-Mode for 2-Mode networks Clumping Starting with network and (random) permutation improve the permutation using clumping algorithm from Murtagh [53, page 11-16]. 1-Mode for ordinary (1-Mode) networks 2-Mode for 2-Mode networks R-Enumeration Starting with network and (random) permutation find such permutation that enumeration of neighbor vertices are as close to each other as possible.

43 Pajek Manual 41 Functional Composition Let f be a partition or a permutation and g a partition, a permutation, or a vector. The result is new partition, permutation or vector r defined in the following way: r[v] = (f g)[v] = g[f[v]]. Expand Partition Greedy Partition Put vertices with unknown class number (0) in the same class as selected vertices in partition if Input...we can reach selected vertices in at most k-steps. Output...we can come to vertices from selected vertices in at most k-steps. All...Input + Output (forget direction of lines) Classes are joined if one vertex should belong to more classes. Influence Partition Put every vertex with unknown class number (0) in given partition in the same class as is the class of the closest vertex. If several vertices with known class number have the same distance, the highest value is used. Make Multiple Relations Network Transform network to a multiple relation network using a partition: if both endvertices of a line belong to the same class in partition the multiple relations tag will be equal to the class number of endvertices, otherwise it will be 0. Expand Reduction Restore original network from reduced network (hierarchical reduction!) and appropriate hierarchy (result is always undirected network). Identify Identify (reorder and/or join some units). Petri Execute Petri net according to starting marking of places determined by partition. Number of places in network is equal to dimension of partition. Places must be defined first (1..m) then transitions (m + 1..n). What to do if more than one transition can fire? Two possibilities: Random Transition is chosen randomly. Complete Complete tree of all possible transitions is built - result is hierarchy. You can choose the maximum depth of the tree, or execute Petri net as long as possible. Try for example petri2 from the book of Peterson [56, page 21] or petri52 (see Figure 13) data.

44 42 Pajek Manual E3 E4 C3. E5 C4.... C5 M4 M5 M1.. M3. C2 M2.. E2 C1 E1 Figure 13: Petri net

45 Pajek Manual 43 Refine Partition Refine partition according to selected network (reachability). Strong... for directed network. Weak... for undirected network. Leader Partition find clusters of vertices of network inside layers. 3.5 Partition Only Partition is needed as input. Create Constant Partition Create constant partition of selected dimension. Default dimension is the size of selected network (if there is one in memory). Create Random Partition Create random one or two mode partition. Binarize Make binary (0-1) partition from selected partition. Fuse Clusters Fuse selected cluster numbers to a new cluster. Canonical Partition Transform partition to its canonical (unique) form (vertex 1 is always in class 1, the next vertex with smallest number that is not in the same class as vertex 1 is in class 2...). Canonical Partition [Decreasing frequencies] Transform partition to its canonical (unique) form (in class 1 the old class with the highest frequency will be set, in class 2 the old class with the second highest frequency... ). Make Network Generate network from partition. Random Network Generate random network where degrees of vertices are determined using partition. Undirected partition gives degrees of vertices in undirected network. Input partition gives input degrees of vertices. Output partition gives output degrees of vertices. 2-Mode Network Generate 2-mode network: first set consists of vertices (v 1... v n ), second set consists of clusters (c 0... c m ). If vertex i is in cluster j the line from v i to c j is generated. If option Existing Clusters only is selected only clusters containing at least one vertex are generated as vertices in the second set.

46 44 Pajek Manual Make Permutation Make permutation from selected partition. (first all vertices with the lowest class number,...) Make Cluster Transform partition to cluster. Make Hierarchy Transform partition to hierarchy (nested or not). Make Vector Transform partition to vector (V [i] := C[i]). Count, Min-Max Vector info about cluster frequencies and minimum and maximum vector value according to given partition. 3.6 Partitions Operations on two partitions. Two partitions must be selected before performing operations. Extract second from first Extract from first partition vertices that satisfy criterion (are on specified interval) determined by second partition. This operation is useful when we have partition that actually saves some information about vertices (for example gender). When you get (extract) some smaller part of the network (for example vertices that are on distances less than 3 from selected vertex), information about gender would be lost without performing the same operation (extraction) on partition. Add Partitions Add two partitions (useful for example when combining Input and Output neighbors in acyclic networks). Min (C1, C2) Minimum of two partitions. Max (C1, C2) Maximum of two partitions. Fuse Partitions Fuse two partitions add second to the end of the first (useful for 2-mode networks). Expand Expand partition to higher (original) dimension. First according to Second (Shrink) Expand first partition according to shrinking determined by second partition. Insert First into Second according to Third (Extract) The current partition was obtained by extracting selected classes defined by the second partition from the first partition. This sub-partition was modified. Using this operation we can insert this modified sub-partition back to the first partition.

47 Pajek Manual 45 Intersection of selected partitions. Cover with Let p be a partition, b a binary partition, and c selected cluster number. Result is new partition q determined in the following way: if b(v) = 0 then q(v) = p(v) else q(v) = c. Merge Let p and q be partitions and b a binary partition. Result is new partition s determined in the following way: if b(v) = 0 then s(v) = p(v) else s(v) = q(v). Make Random Network generate random network with input degrees determined by the first and output degrees by the second partition. Info Bivariate statistical measures between selected partitions: Cramer s V, Rajski, Adjusted Rand Index Report contingency table, compute Cramer s V, Rajski coefficients, and Adjusted Rand Index. Spearman Rank correlation coefficient. 3.7 Vector Operations using vector. Create Constant Vector Create constant vector (vector with all values equal to selected value) of selected dimension. Default dimension is the size of selected network (if there is one in memory). Extract Subvector Extract subvector from given vector - criterion is class in the selected partition. Shrink Vector Shrink vector values according to clusters of partition to new vector adjusting vector to shrunken network. When shrinking several values to one value, sum of values, mean, min, max or median value can be used. Make Partition Convert vector to partition: by Intervals according to selected dividing numbers in vector vertices get appropriate class numbers. Intervals can be given by: First Threshold and Step Select first threshold and step in which to increase threshold.

48 46 Pajek Manual Selected Thresholds Select all thresholds or number of classes (#) in advance. by Truncating (Abs) partition is absolute and truncated vector. Make Permutation Convert vector to permutation - sorting permutation. Make 2-Mode Network Convert vector to 2-mode network (row or col). Transform Transformations of given vector: Multiply by a constant. Add Constant to vector values. Absolute values of its elements. Absolute + Sqrt square root of its absolute components. Truncate truncated vector. Exp exponential of vector. Ln natural logarithm of vector. Power selected power of vector. Normalize 3.8 Vectors Sum normalize so that the sum of elements is 1. Max normalize so that the maximum element will have value 1. Standardize normalize so that arithmetic mean will be 0 and standard deviation 1. Invert inverse values of vector (exception is that 0 stays 0). Operations on two vectors. Two vectors must be selected before performing operations. Add Vectors sum of selected vectors. Subtract Second from First difference of selected vectors. Multiply Vectors product of selected vectors. Divide First by Second division of selected vectors. Linear Regression fit the two vectors using linear regression. Results are: regression line, linear estimates of second vector and corresponding errors.

49 Pajek Manual 47 Min (V1, V2) smaller elements in selected vectors. Max (V1, V2) bigger elements in selected vectors. Fuse Vectors fusion of vectors. Transform two vectors to another two vectors: Cartesian Polar First vector must contain x-coordinates second y-coordinates. Results are: vector containing polar radius and vector containing polar angles in degrees. Polar Cartesian First vector must contain polar radius second polar angles in degrees. Results are: vector containing x-coordinates and vector containing y-coordinates. Results can be (de)normalized to enable direct use in Draw window. Info Pearson correlation coefficient between selected vectors. 3.9 Permutation Only permutation is needed as input. Identity Create identity permutation of selected dimension. Default dimension is the size of selected network (if there is one in memory). Random Create random permutation of selected dimension. Default dimension is the size of selected network (if there is one in memory). Random 2-Mode Create random permutation of selected dimension and number of vertices in the first subset of 2-mode network. Default dimension is the size of selected network and number of vertices in the first subset (if there is network and corresponding partition in memory). Inverse Create inverse permutation of selected permutation. Mirror Create mirroring permutation of selected permutation (sort in opposite direction). Make Partition Create partition into selected number of clusters from given permutation. Make Vector Transform permutation to vector.

50 48 Pajek Manual 3.10 Permutations Operation on two permutations. Fuse Permuations Fuse two permutations add second to the end of the first (useful for 2-mode networks) Cluster Only cluster (and partition) is needed as input. Create Empty Cluster Create cluster without vertices. Create Complete Cluster Create cluster with values 1..n. Make Partition Transform cluster to partition. Binarize Partition Binarize partition according to cluster - make binary partition of the same dimension as the given partition, vertices that are in cluster numbers determined by the cluster will go to class 1 other to class 0. This allows noncontiguous ranges to be selected (other choices in Pajek need contiguous ranges). Note the exception: In this case cluster represents set of cluster numbers and not set of vertices numbers Hierarchy Only hierarchy is needed as input. Extract Cluster Extract cluster from hierarchy - the cluster is whole subtree of selected node in hierarchy. Make Network Converts hierarchy to network (use it for example to draw hierarchy drawing by layers). Closed nodes are also taken into account. Make Partition Converts hierarchy to partition (according to closed nodes). Make Permutation Converts hierarchy to permutation. Export as Dendrogram to EPS Draw dendrogram of hierarchy in EPS. Works for binary hierarchies only. Dissimilarities must be stored in names of nodes of hierarchy between [ and ]. These are obtained automatically when obtaining hierarchies using hierarchical clustering or clustering with relational constraint.

51 Pajek Manual Options Read - Write Threshold Value of line must be higher (absolutely) than the given threshold to generate line between two vertices. Max. vertices to draw Maximum number of vertices in network to allow drawing (to prevent long waiting). Large Network (Vertices) Select the threshold number of vertices that defines very large network. For such networks Pajek (for some operations), asks if the old network can be destroyed and replaced by the network that is obtained as a result. In this way we can spare a lot of memory. Read - Save vertices labels? Read / Save also labels, coordinates, and other descriptions of vertices or not. If vertices labels are not read (recommended if network is very large and vertices labels are long) they can be imported later from input file using Net/Transform/Add/Vertex Labels from File. Save coordinates of vertices? Save coordinates of vertices to network file (or not). Save complete vertex description? When saving network to output file for each vertex complete description will be written, even if consequent vertices have the same descriptions (e.g. shapes, time intervals...). Check equality of vertex descriptions by reading? Enables users to speed up reading large network files according to descriptions of vertices: Check this option to save space when exactly the same descriptions of vertices are repeated often (e.g. shapes of vertices). Uncheck this option to save time when there are several different desctriptions of vertices in input file (e.g. time stamps in temporal networks). Check equality of line descriptions by reading? Enables users to speed up reading large network files according to descriptions of lines: Check this option to save space when exactly the same descriptions of lines are repeated often (e.g. line pattern Dots/Solid). Uncheck this option to save time when there are several different desctriptions of lines in input file (e.g. time stamps in temporal networks). Auto Report? Automatically report all text results to file rep1.rep. Ore: Different relations for male and female links When reading genealogy as Ore graph generate 2 different types of arcs: arc with relation number 1 (also value 1) represents (god)father to child relation,

52 50 Pajek Manual arc with with relation number 2 (also value 2) represents (god)mother to child relation. Ore: Generate Godparent relation When reading genealogy as Ore graph generate also godparent relation (relation number 4) or godfather (relation number 4) and godmother (relation number 5) relations. GEDCOM Pgraph Use pgraph format (nodes are couples or individuals) when reading genealogies (D. R. White), otherwise nodes are only individuals. Bipartite Pgraph Generate bipartite pgraph that has squares for marriages, triangles and circles for individuals. Pgraph+labels Attach also labels of lines to pgraph, when reading GEDCOM file. x / 0 = Specify the result when dividing nonzero value with 0. 0 / 0 = Specify the result when dividing 0 with 0. Ignore Missing Values in Info/Vector? When computing descriptive statistics on Vector treat missing values (values larger that ) as valid numbers. Save Files as Unicode UTF8 with BOM? Instead of saving to ASCII files save files as Unicode UTF8 files. Select Font Select Proportional Font Select proportional font for displaying Unicode characters (e.g. in Draw window). Select Monospaced Font Select non-proportional font for displaying Unicode characters (e.g. in Report window). Default Fonts Use MS Sans Serif for proportional, and Courier New for monospaced font. Main Window Colors - Select color for panels, color for drop-down menus and color for font used in Pajek main window. Blockmodel Select type of blockmodel for shrinking. Possibilities are: 0..Min Number of Links 1..Null 2..Complete

53 Pajek Manual 51 Figure 14: Generalized block types

54 52 Pajek Manual 3..Row-Dominant 4..Col-Dominant 5..Row-Regular 6..Col-Regular 7..Regular 8..Row-Functional 9..Col-Functional 10..Degree Density Look in Batagelj [7] and Doreian, Batagelj, Ferligoj [35]. Ini File Load Use selected configuration of Pajek which is stored in the file (*.ini). Save Save the current configuration of Pajek into a file (*.ini). Use Old Style Dialogs If Windows 7 have problems with opening/saving files check this option Info Network Information about network General General information about network number of vertices number of arcs, edges and loops density of lines average degree sort lines according to their values (ascending or descending) to find the most/least important lines. Line Values Frequency distribution of line values. Indices Different indices on network (chemical and genealogical). Triadic Census Number of different triads in network. See book of Faust and Wasserman [64] and Figure 15 on page 55. Multiple Relations General information about multiple relations network

55 Pajek Manual 53 number of relations number of arcs, edges and total number of lines for each relation Vertex Label -> Vertex Number Find vertex number by giving (part of) its label, or find vertex label for given vertex number. Partition General information about partition. Sort vertices according to their class numbers (ascending or descending) to see the most important vertices. Frequency distribution of class numbers. Average, median and standard deviation of class numbers are also given. Hierarchy General information about hierarchy. Operation is possible only if node numbers are integers. It returns number of vertices in nodes of hierarchy (on first level). Vector General information about vector: Vertices sorted according to their values, average, median, standard deviation and frequency distribution of vector values into given number of classes (# number of classes or selected dividing values can be given). Memory Available memory. Not very accurate. About Information about Pajek version, authors, copyrights Tools R SPSS Send to R Call statistical package R [57] with one vector/network, vectors/networks selected by cluster or all currently available vectors and/or networks. Locate R locate position of statistical program R (Rgui.exe or Rterm.exe) on the disk. Send to SPSS Call statistical package SPSS with one partition, vector or network, partitions/vectors selected by cluster or all currently available partitions and vectors. Locate SPSS locate position of statistical program SPSS (runsyntx.exe) on the disk.

56 54 Pajek Manual Export to Tab Delimited File Export Networks, Partitions, and Vectors to tab delimited file. This file can then be imported to other programs, like statistical packages, Excel... Web Browser Select which web browser to open when clicking on vertex with Shift and Right mouse button. Add Program add new executable program with specified parameters to the tools menu. Edit Parameters edit parameters of selected external program. Remove Program remove selected external program from the tools menu.

57 Pajek Manual D 5-021U 6-021C 7-111D 8-111U 9-030T C D U C Figure 15: All different triads.

58 56 Pajek Draw window 4 Draw Window Tools 4.1 Main Window Draw Tool Draw - Draw Network. A new window is open, where a new menu appears. You can edit network by hand (move vertices using left mouse button), select a part of the picture (using right mouse button and select the area), edit lines that belong to selected vertex by clicking on vertex using right mouse button, spin picture using keys X, Y, Z, S, x, y, z, s. Description of Draw window menu: Draw-Partition Similar to Draw. Colors of vertices represent the classes in selected partition. Additionally you can put selected vertex or selected vertices into given class in partition (classes are shown using different colors) by clicking on middle mouse button (or Shift+left button) (increment class), or together with Alt (or Alt+left button) - decrement class number. In Figure 20 on page 90 you can see which color represents selected class. Some additional menu items that were already described appeared (you can draw network according to layers from partition and optimise energy using fixed vertices determined using partition). It is also possible to move the selected class (by clicking close to vertex from that partition). Draw-Vector Sizes of vertices are determined using selected vector. Draw-2Vectors Sizes of vertices are determined using selected two vectors (first for width second for height). Draw-Partition-Vector Colors of vertices are determined using selected partition, sizes of vertices are determined using selected vector. Draw-Partition-2Vectors Colors of vertices are determined using selected partition, sizes of vertices are determined using selected two vectors (first for width second for height). Draw-SelectAll Create null partition and draw network using it. 4.2 Layout Generate layout of the network. Circular Position vertices on circle 1. Original in order determined by the network.

59 Pajek Draw window using Permutation in order determined by current permutation. 3. using Partition create separate circles for clusters in selected partition. Center of the circle is determined by the arithmetic mean of positions of vertices in the cluster. 4. Random in random order. Energy Automatic layout generation. 1. Kamada-Kawai algorithm for automatic layout generation in the plane. (a) Free Every position in the plane is possible. (b) Separate Components Optimize each component separatelly and tile components at the end. (c) Fix first and last First and last vertex are fixed in opposite corners. (d) Fix one vertex in the middle - Select vertex which will be fixed in the middle of the picture. (e) Selected group only Only selected part of the picture is taking into account during optimisation. (f) Fix selected vertices - Selected vertices (from partition) are fixed on given positions). This item is visible only if Draw partition is active. 2. Fruchterman Reingold another algorithm for automatic layout generation (faster than Kamada-Kawai). (a) 2D optimisation in plane. (b) 3D optimisation in space. (c) Factor Input factor for optimal distance among vertices when using Fruchterman Reingold optimisation. 3. Starting positions for energy drawing (random, circular, given positions on plane xy, given z coordinates). EigenValues Drawing using eigenvalues/eigenvectors (Lanczos algorithm). Values of lines can be taken into account or not Select 2 or 3 eigenvalues and algorithm will compute corresponding eigenvectors. Eigenvalues may be multiple so there are many possibilities. Some examples (a) compute 3 eigenvectors that correspond to the first eigenvalue

60 58 Pajek Draw window Figure 16: VRML display of layout determined by eigenvectors (b) compute 2 eigenvectors that correspond to the first eigenvalue and 1 that correspond to the second (c) compute 1 eigenvector that correspond to the first eigenvalue and 2 that correspond to the second (d) compute 1 eigenvector that correspond to the first, 1 that correspond to the second and 1 that correspond to the third eigenvalue (e) 1 1 compute 2 eigenvectors that correspond to the first eigenvalue (2D picture) Tile Components Tile weakly connected components in a plane. 4.3 Layers Visible only if Draw partition is active. Draw in layers according to partition. Type of Layout Select type of picture (2D layers in y direction, or 3D layers in z direction). According to that, appropriate menu appears. In y direction Draw vertices in layers (y coordinate) inside layers draw vertices centered-equidistantly (x coordinate), z coordinate is 0.5 for all vertices.

61 Pajek Draw window 59 In y direction+random in x Same as first option, only vertices are put on layers in random order not according to vertex numbers. In z direction Draw layers in z direction, live x and y coordinates as they are. In z direction + random in xy Draw layers in z direction, x and y coordinates get random values Averaging x coordinate Use it after vertices are put on layers in 2D. Iterativelly compute average x-coordinate of all neighbors and normalize. Good approximation of global picture, but vertices are put to close to each other. Use it on all vertices or only on selected one. Averaging x and y coordinates Use it after vertices are put on layers in 3D. Iterativelly compute average x and y coordinates of all neighbors and normalize. Good approximation of global picture, but vertices are put to close to each other. Use it on all vertices or only on selected one. Tile in x direction After averaging x coordinate vertices are put to close to each other, so using this option vertices are repositioned to minimal distance described in resolution. Tile in xy plane Same as previous, only this option is used when drawing 3D pictures. Optimize layers in x direction Optimize layout in layers using minimization of the total length of lines. 1. Forward go from first to last layer. In the current layer optimize only layers having numbers equal or one smaller as the current layer number. 2. Backward go from last to first layer. In the current layer optimize only layers having numbers equal or one larger as the current layer number. 3. Complete go from first to last layer. In the current layer optimize only layers having numbers equal or one smaller or one larger as the current layer number. Optimize layers in xy plane Same as previous, only this option is used when drawing 3D pictures.

62 60 Pajek Draw window Resolution How many additional positions are available on layers. Works only if Pgraph is selected. The higher the resolution, the better is the result of optimization, but also slower. 4.4 GraphOnly Show complete picture without labels and arrows. 4.5 Previous Draw previous network, and/or partition and/or vector which are loaded in Pajek (depending on selection in Options/PreviousNext/Apply to). 4.6 Redraw Redraw network. 4.7 Next Draw next network, and/or partition and/or vector which are loaded in Pajek (depending on selection in Options/ PreviousNext/ Apply to). 4.8 ZoomOut Zoom out (visible only when zooming in layout). 4.9 Options Additional options for picture layout. Transform Transformations of picture. 1. Fit area (a) max(x), max(y), max(z) Draw picture as big as possible to fit area (resize each coordinate to fit in picture independently). (b) max(x,y,z) Resize to fit in area but keep real proportions (resize according to largest distance in all three coordinates, e.g. molecule). 2. Resize Resize picture (or selected part of it) in all three directions for selected factor.

63 Pajek Draw window Translate Translate picture (or selected part of it) in space. 4. Reflect y axis Reflect picture (or selected part of it) around y axis. 5. Rotate 2D Rotate picture (or selected part of it) in xy plane. 6. Fisheye Fisheye transformation (cartesian or polar) of picture. If no vertex is selected the middle point of picture (or selected part of it) is used as focus, otherwise first selected vertex will be used as focus. Values of lines Meaning of values of lines during energy drawing or eigenvectors computing (no meaning, similarities, dissimilarities (distances)). Mark vertices using Labels can be marked using 1. labels 2. numbers 3. partition clusters (if partition of the same size is also selected) 4. vector values (if vector of the same size is also selected) 5. without labels 6. without labels and arrows 7. cluster only only vertices belonging to the current Cluster are labeled in the layout. Lines Select the way the lines are drawn: 1. Draw Lines (a) Edges draw edges or not (b) Arcs draw arcs or not (c) Relations draw all lines (leave empty string) or just lines belonging to selected relations, e.g. 1-3,6, Mark Lines (a) No do not mark lines (b) with Labels (c) with Values 3. Different Widths if checked, the width of the lines will be determined by their line values. 4. GreyScale if checked, the color in grayscale of lines will be determined by their line values.

64 62 Pajek Draw window Size Determine size of vertices, width of vertices border, width of lines, size of arrows, size of font or turn them to default values. Sizes of vertices can be set to autosize (0-average). They can be read from input file (x fact and y fact) or determined by one vector (currently selected) or the two currently selected vectors. FontSize can be proportional to values stored in third partition. Colors Determine color of background, vertices, border of vertices, edges, arcs and font (of vertices labels and lines labels) or turn them to default values. Colors of edges/arcs can represent relation number of lines. You can select which color should represent given class using Relation Colors. You can also use colors of vertices (ic Red), border of vertices (bc Blue), arcs and edges (c Green) as defined on input file (see Figure 19, page 89). FontColor can be determined by values stored in second partition. Additionally to predefined colors it is possible to specify colors using RGB (e.g. RGBFF0000, or RGB(1,0,0)) and CMYK (e.g. CMYK00FF0000, or CMYK(0,1,0,0)) format. Note that there should be no spaces in string defining a color. Example NET file: *Vertices 9 1 "a" ic Pink bc Black 2 "b" ic CMYK(0,0,1,0.0) bc CMYKFF "c" ic Cyan bc Yellow 4 "d" ic Purple bc Orange 5 "e" ic Orange bc Brown 6 "f" ic Magenta bc Green 7 "g" ic Brown bc Magenta 8 "h" ic RGB(1,0,0) bc Blue 9 "i" ic Green bc Magenta *Arcs c RGB0000FF c Red c Black c Yellow c Gray c Cyan c Magenta c Purple c Brown Layout Layout options 1. Redraw Redraw whole network during or/and after moving of selected vertex, and/or redraw if draw window is paint. 2. Real xy proportions The draw window has always square shape or not.

65 Pajek Draw window Arrows in the Middle Draw arrows in the middle of lines not at terminal vertices. 4. Size of vertex 0 How to handle vertices of size 0 (size of vertex is determined in input file or using vector)? (a) Hide vertex vertices of size 0 are shown or not. (b) Hide attached lines lines with one end in vertex of size 0 are shown or not. 5. Decimal Places How many decimal places to use when marking vertices using vectors. 6. Show SubLabel Select the position of the vertices sublabel that is shown in network layouts. ScrollBar On/Off Show/Hide the scrollbars in the top left corner of Draw window. When part of picture is selected, scrollbar is used for moving. When whole picture is selected, scrollbar is used for spinning (like pressing keys X, Y, Z, x, y, z spinning around axis defined by the key, and S spin around selected normal) Interrupt Interrupt period during optimisation (stop every? second, or not) Previous/Next Select parameters when using Previous or Next commands for drawing sequence of networks in draw window: 1. Max. number How many networks to show in sequence. If the number is higher than number of existing networks the sequence will stop earlier. 2. Seconds to wait Seconds to wait between the two layouts. 3. Optimize Layouts Optimize the current layout or not. If the sequence of networks is obtained from the same network, it is useful to choose Energy/Starting Positions/ Given xy to start optimization with existing coordinates. (a) Kamada-Kawai Optimize the current layout using Kamada-Kawai algorithm. (b) 2D Frucht. Rein. Optimize the current layout using 2D Frucht. Rein. algorithm. (c) 3D Frucht. Rein. Optimize the current layout using 3D Frucht. Rein. algorithm. (d) No Do not optimize the layout, just show the picture.

66 64 Pajek Draw window 4. Apply to Which object (Network, Partition, Vector) will change when Previous or Next is selected: (a) Network The previous / next network in memory is drawn. If Draw-Partition is selected and the new network matches in dimension with selected partition, the same partition will determine colors of vertices of the new network. If Draw-Vector is selected and the new network matches in dimension with selected vector, the same vector will determine sizes of vertices of the new network. Option can be used to show several networks of equal size using the same partition/vector. (b) Partition The previous / next partition in memory is selected. If Draw-Partition is selected: The same network is drawn using previous / next partition (network and partition must match in size). Option can be used to show several partitions of selected network. (c) Vector The previous / next vector in memory is selected. If Draw-Vector is selected: The same network is drawn using previous / next vector (network and vector must match in size). Option can be used to show several vectors of selected network. By checking several objects (Network, Partition, Vector) at the same time previous / next networks will be drawn using previous / next partitions and (or) vectors at the same time. All consequent selected objects must match in size. Select all Select all vertices in window (then possible to put vertices in given class) Export Export layout of the network to one of the following two or three dimensional formats: 2D two dimensional exports: EPS/PS Export to EPS format (with or without Clip, or WYSIWYG [What You See Is What You Get exported EPS picture is similar to picture in Draw window except that colors are Black/White or color determined by partition]). PS Export to PS format (similar to EPS but without header).

67 Pajek Draw window 65 SVG Export to SVG (Scalable Vector Graphics) format. Additional controls over layout can be included in SVG or HTML. The plugin for examining layouts can be obtained from Adobe [1]. Linear or radial gradients (continuously smooth color transitions from one color to another) can be selected as well up to three background colors can be selected in Export/Options window. 1. General Export to SVG without possibility to choose parts of the picture. 2. Labels/Arcs/Edges Export with possibility to turn labels, arcs and/or edges on/off. 3. Partition Export to SVG using one or two partitions. One partition is used by default: the same partition determines classes and colors. But, if two partitions are defined by Partitions menu, first partition will determine classes, the second colors of vertices. (a) Classes User can turn selected classes and lines among classes on/off. (b) Classes with semi-lines User can turn selected classes on/off. Lines among classes are drawn as semi-lines. (c) Nested Classes Upper classes are nested in lower whenever selected class is turned on all higher classes are turned on too, and all lower classes are turned off (suitable for showing cores, for example). 4. Line Values Export to SVG using values of lines. Threshold values or number of classes must be given. If you input #n, n classes of equal size will be generated. According to obtained thresholds, subsets of lines (and incident vertices) are defined and can be turned on/off inside web browser. (a) Classes User can turn lines of selected value and incident vertices on/off. (b) Nested classes User can turn lines of selected value or higher (lower) and incident vertices on (off). (c) Options Additional options to emphasize the values of lines using some visual properties. Different Colors Subsets of lines are drawn using colors which are used for partitioning vertices too. Using GreyScale The darkness of a line corresponds to its value. Different Widths The width of a line corresponds to its value.

68 66 Pajek Draw window 5. Multiple Relations Network Export multiple relations network to SVG. User can show/hide one or more relations in the layout. 6. Current and all Subsequent If checked the current network and all subsequent networks will be exported to SVG. For each network separate html file is generated. Files are given the following names: file001.htm, file002.htm,... Generated html files get additional links (Previous/Next) to transition among them. If also next partitions / vectors fit in dimension to dimension of networks, partitions will determine color of vertices, vectors will determine sizes of vertices. Subsequent is applied to any combination of [Network, Partition, Vector] (one of the three objects only, any pair of them or all three of them) according to selection in Options/Previous/Next/ Apply to in Draw window. Bitmap Export to Windows bitmap (bmp) format. 3D three dimensional exports: X3D Export to X3D (XML based 3D computer graphics, the successor of VRML) format. Kinemages Export to Kinemages format with balls or labels. You need Mage or King viewer to watch it. A free copy of the Mage software can be downloaded from its site [58]. 1. Current Network Only Export only current network to Kinemages. Two partitions defined by Partitions menu can be used - one for generations one for colors. 2. Current and all Subsequent Export current network and all subsequent networks (use commands KINEMAGE/Next or Ctrl N in Mage). If also next partitions/vectors fit in dimension to dimension of networks, partitions will determine color of vertices, vectors will determine sizes of vertices. Subsequent is applied to any combination of [Network, Partition, Vector] (one of the three objects only, any pair of them or all three of them) according to selection in Options/Previous/Next/ Apply to in Draw window. 3. Multiple Relations Network Export with possibility to hide / show selected relations. VRML Export to VRML (Virtual Reality) format. For examining it you need a VRML viewer such as Cortona [29] or (older) Cosmo player [30].

69 Pajek Draw window 67 MDL MOL file Export to MDL Molfile format. You need Chime plugin (Chemscape Chime) for Netscape to explore it [51]. Options EPS, SVG, X3D and VRML default options (see section on Exports to EPS/SVG/X3D/VRML). Append to Pajek project file Add current network to the end of selected project file (used by program PajekToSvgAnim) Spin Select file Select project file. Append Append to project file. Spin around Spin network around selected normal. Perspective Distant vertices are drawn smaller (or not). Normal Normal vector to spin around. Step in degrees Step in degrees when showing rotation Move Give additional constraints on hand vertex moving: Fix Fix (do not allow) moving in x or y direction, or do not allow changing distance from center (circulating). Grid Define (x, y) positions on grid, these become feasible positions for vertices during moving by hand. Circle Define (x, y) positions on concentric circles, these become feasible positions for vertices during moving by hand. Grasp Determine which additional vertices are moving when clicking with left mouse button close to vertex in given class. Vertices that will be moved are in the: 1. Closest Class Only 2. Closest Class and Higher 3. Closest Class and Lower

70 68 Pajek Draw window 4.13 Info Select aesthetic properties of the current layout to compute: Closest Vertices Smallest Angle Shortest/Longest Line Number of crossings if lines Vertex Closest to Line All Properties Remember that coordinates of vertices must be between 0 and 1!

71 Exports to EPS/SVG/X3D/VRML 69 5 Exports to EPS/SVG/X3D/VRML 5.1 Defaults If you have no labels in Draw window when you call Export there will also be no labels in EPS/SVG picture, otherwise numbers/labels like in Draw window will be shown. If you are looking at the picture using Draw/Partition, the same colors will be automatically used in EPS/SVG picture too. 5.2 Parameters in EPS, SVG, X3D, and VRML Defaults Window Window is divided into 5 frames, two on the left and three on the right. Note that settings you made in this window are overwritten if the parameters are specified in Pajek input (NET) file. Top frame on the left EPS/SVG Vertex Default This frame defines default drawing of vertices when we export layouts to EPS and SVG: Interior Color interior color of vertices (see Figure 19, page 89). If drawing using Partition is used, partition colors will overwrite the specified interior color. Border Color color of the borderline of vertices. Label Color color of label of vertices. Border Width width of the borderline of vertices. Label Position: Radius /Angle position where the label will be displayed: Radius distance of beginning of vertex label from vertex center first polar parameter. Angle position of vertex label in degrees - second polar parameter (0..360). If radius is equal to 0, the label will be centered, otherwise it will be left aligned on the specified position. Fontsize size of font of vertices labels.

72 70 Exports to EPS/SVG/X3D/VRML Label Angle angle in which vertex labels will be displayed: if angle is smaller than 360, it means relative according to horizontal line (0 horizontally); otherwise it is relative to center of the layout (360 concentrically). The latter is useful when all vertices are drawn in concentric circle(s) using Layout/Circular. x/y Ratio ratio between size of vertex in x and y direction (e.g., value 1 for circle, value larger/smaller than 1 for ellipse). Shape default shape of vertex (ellipse, box, or diamond). Shapes file default shapes file, double click to change it. Export options overwrite shapes file if checked: options selected in this window overwrite options defined in shapes file, otherwise default values for each shape are defined in selected shapes file. Bottom frame on the left EPS/SVG Line Default This frame defines default drawing of lines when we export layouts to EPS and SVG: Edge Color color of edges. Edge Width width of edges. Arc Color color of arcs. Arc Width width of arcs. Pattern pattern for drawing lines (Solid or Dots). Arrow Size size of arrows. Arrow Position distance of arrow from terminal vertex: If distance is between 0 and 1, it means relative distance according to arc length, e.g.: 0 arrow touching the terminal vertex; 0.5 arrow in the middle of the arc. If distance is larger than 1 it means absolute distance from the terminal vertex (this is useful if you want to have all arcs on the same distance from terminal vertex, regardless of arcs length) Label Color color of line labels. Label Angle angle in which label of line will be displayed: if angle is smaller than 360, it means relative to direction of line (0 parallel to line); otherwise it is relative to horizontal line (360 horizontally).

73 Exports to EPS/SVG/X3D/VRML 71 Label Position position of the center of line label position is point on the line - distance of the center of line label from terminal vertex (see also Arrow Position) Fontsize size of font of line labels. Label Position: Radius /Angle position where the center of line label will be displayed according (relative) to Label Position: Radius distance of center of line label from point defined by Label Position - first polar parameter. Angle position of label in degrees - second polar parameter (0..360). Only straight lines drawing bidireced and multiple lines without curves. Top frame on the right This frame defines some additional defaults when we export layouts to EPS, SVG, X3D, or VRML: EPS, SVG, X3D, VRML Size of Vertices default size of vertices when exporting to X3D/VRML (valid for EPS and SVG exports as well). X3D/VRML Size of Lines width of lines in X3D/VRML. EPS: Use RGB colors instead of CMYK in description of colors in EPS file RGB is used (default is CMYK). SVG: 3D Effect on Vertices if selected, gradient will be applied to get 3D look of vertices. Middle frame on the right Background Colors This frame defines background color when exporting to EPS/SVG/X3D/VRML and gradients (continuously smooth color transitions from one color to another) when exporting layouts to SVG/X3D: Bckg. Color 1 Background color for layout in EPS/SVG/X3D/VRML. Bckg. Color 2 the second color for SVG/X3D export. No means without second color, otherwise selected gradient will be used. Bckg. Color 3 the third color for SVG/X3D export gradient. Gradients type of gradient to use in SVG (No, Linear, or Radial). In X3D only Radial gradient is supported.

74 72 Exports to EPS/SVG/X3D/VRML Bottom frame on the right This frame defines some additional defaults when we export layouts to EPS: Left, Right, Top, Bottom additional border around layout when exporting to EPS (only when EPS Clip format is selected) Border Color color of borderline of layout. It is used for export to SVG as well. No means without borderline. Border Radius radius of borderline of layout (if greater than 0 it means oval instead of rectangle). Border Width width of borderline of layout. Figure 17: Spider / Križevec; Photo: Stana.

75 Exports to EPS/SVG/X3D/VRML Exporting pictures to EPS/SVG defining parameters in input file Definition of Network (and its picture) on Input ASCII File For every vertex and line we can specify in details how it should be drawn (colors, shapes, sizes, patterns, rotations, widths...). A kind of standardized language is used for describing networks. The following reserved words are used: 1. * Vertices n definition of vertices follows. n is number of vertices. Each vertex is described using following description line: vertex num label [x y z] [shape] [changes of default parameters] Explanation: vertex num vertex number (1, 2, 3... n) label if label starts with character A..Z or 0..9 first blank determines end of the label (e.g., vertex1), labels consisting of more words must be enclosed in pair of special characters (e.g., vertex 1 ) x, y, z coordinates of vertex (between 0 and 1) shape shape of object which represents vertex. Shapes are defined in file SHAPES.CFG (ellipse, box, diamond, triangle, cross, empty) Description of parameters in shapes.cfg: SHAPE s s is external name of vertex (used in Pajek network file) sh sh can be ellipse, box, diamond, triangle, cross, empty. This is the name of PostScript procedure that actually draws object (procedure is defined in drawnet.pro). s size default size x fact magnification in x direction y fact magnification in y direction phi rotation in degrees of object in + direction (0..360) r parameter used for rectangle and diamond for describing radius of corners (r = 0 rectangle, r > 0 roundangle) q parameter used for diamonds ratio between top and middle side of diamond (try q 0.01, q 0.5, q 2,...) ic interior color of vertex. See Figure 19, page 89 for the list of possible colors. bc boundary color of vertex

76 74 Exports to EPS/SVG/X3D/VRML bw boundary width of vertex lc label color la label angle in degrees (0..360) lr distance of beginning of vertex label from vertex center (radius first polar parameter) lphi position of label in degrees (0..360) (angle phi second polar parameter) fos font size font PostScript font used for writing labels (Helvetica, Courier,...) HOOKS positions where edges can join the selected shape - according to s size. Three different ways to specify these positions: (a) CART x y positions in Cartesian coordinates (x,y) (b) POLAR r phi positions in polar coordinates, phi is positive angle (0..360) (c) CIRC r phi1 iteration of positions in polar coordinates r radius, phi = k phi1, k = 1, 2,..; k phi1 360 Default values can be changed for each vertex in definition line, example: 1 vertex one box ic White fos 20 Explanation: White box will represent vertex 1, label (vertex one) will be displayed using font size * Arcs (or * Edges) definition of arcs (edges). Format: v1 v2 value [additional parameters] Explanation: v1 initial vertex number v2 terminal vertex number value value of arc from v1 to v2 These three parameters must always be present. If no other parameter is specified, the default arc will be black, straight, solid arc with following exceptions: if value is negative, dotted line will be used instead of solid, if arc is a loop (arc to itself) bezier loop will be drawn, if bidirected arc exists two curved bezier arcs will be drawn.

77 Exports to EPS/SVG/X3D/VRML 75 Arrow will be drawn at the end of the edge (at terminal vertex). As we mentioned, hooks are used to specify exact position where line joins vertices. Additional parameters: w width of line c color of line p pattern of line (Solid, Dots) s size of arrow a type (shape) of arrow (A or B) ap position of arrow ap = 0 arrow at terminal vertex 0 < ap 1 proportional distance from terminal vertex (according to line length) ap > 1 absolute distance l line label (e.g. line 1 2 ) lp label position (look at ap) lr label radius (position of center text from point on edge ) lphi label radius (angle of center text according to point on edge ) (lr and lphi are polar coordinates) lc label color la label angle (0 < la < 360 relative to edge, la 360 absolute angle according to x axis) fos font size of label font PostScript font used for writing labels (Helvetica, Courier,...) h1 hook at initial vertex (0 center, -1 the closest, 1, 2.. user defined) h2 hook at terminal vertex a1 angle at initial vertex (Bezier) k1 velocity at initial vertex (Bezier) k2 velocity at terminal vertex (Bezier) a2 angle at terminal vertex (Bezier)

78 76 Exports to EPS/SVG/X3D/VRML Special shapes of lines can be defined using combinations of alpha1, k1, alpha2, k2: alpha1 = alpha2 = 0, k1 0, k2 0 straight line (default) alpha1 = alpha2 = 0, k1 = 1, k2 > 0 oval edge with radius k2 (measure of radius is absolute as explained above) alpha1 = alpha2 = 0, k1 = 1, k2 < 0 second possible oval edge with radius k2 alpha1 = alpha2 = 0, k1 = 2, k2 > 0 circular arc with radius k2 in positive direction alpha1 = alpha2 = 0, k1 = 2, k2 < 0 second possible circular arc with radius k2 in positive direction alpha1 = alpha2 = 0, k1 = 3, k2 > 0 circular arc with radius k2 in negative direction alpha1 = alpha2 = 0, k1 = 3, k2 < 0 second possible circular arc with radius k2 in negative direction alpha1 = alpha2 = 0, k1 = 4 double edge alpha1 or alpha2 0, k1 > 0, k2 > 0 Bezier curve (if alpha1 and alpha2 have different signs line goes from one to another side of straight line connecting both vertices, if alpha1 and alpha2 have the same sign line stays on the same side of straight line connecting both vertices) 3. * Edges definition of edges. The same parameters as for arcs can be used, except that type (a), size (s) and position (ap) have no meaning for edges. PS and EPS PS Export to Postscript (.PS) without header (drawnet.pro). Use it to spare space, if you have many pictures and your word processor enables you to define the header separately (like in L A TEX). EPS Export into file with Encapsulated PostScript description (.EPS). Drawnet.pro is already inserted in the beginning. The picture is complete, you can include it into text, make it bigger or smaller (without losing quality), rotate it, print it on Postscript printer, view it with GhostScript viewer, convert it to PDF, JPG...

79 Exports to EPS/SVG/X3D/VRML 77 An example *Vertices 4 1 "ellipse" ellipse x_fact 5 y_fact 3 fos 20 ic LightYellow lc Red 2 "box" box x_fact 5 y_fact 3 fos 20 ic LightCyan lc Blue 3 "diamond" diamond x_fact 7 y_fact 3 fos 20 ic LightGreen lc Black 4 "triangle" triangle x_fact 9 y_fact 3 fos 20 ic LightOrange lc Brown lr 30 lphi 200 *Arcs h2 0 w 3 c Blue s 2 a1-130 k1 0.6 a2-130 k2 0.6 ap 0.25 l "Bezier loop" lc OliveGreen fos 20 lr 13 lp 0.5 la h2 1 a1 120 k1 1 a2 10 k2 0.8 ap 0 l "Bezier arc" lphi 270 la 180 lr 13 lp h2-1 a1 40 k1 2 a2-30 k2 0.8 ap 0 l "Bezier arc" lphi 90 la 0 lp l "Straight dotted arc" p Dots c Red *Edges l "Straight edge" lp l "Straight edge" You have to set some options in Pajek s draw window: for example Options / Lines / Mark Lines / with labels to activate the display of line labels. Bezier loop Bezier arc ellipse box Bezier arc Straight dotted arc Straight edge diamond Straight edge triangle Figure 18: Example picture 5.4 Using Unicode in Pajek s pictures Sometimes we would like to use in Pajek s picture some characters from unusual alphabets (special symbols, Cyrillic, Greek, Arabic, Chinese,...). They are available in Unicode. Pajek supports Unicode UTF8 files with BOM from version 2.00 on.

80 78 Exports to EPS/SVG/X3D/VRML Some hints on using Unicode in Pajek: Using the free Unicode editor BabelPad we prepare the Unicode version of the *.NET file. Use the U tool (Unicode table) to enter the Unicode characters. Using the BabelPad option File/Save as/utf8 and checking Byte Order Mark you save network in a Unicode file. You can use also File/Save as/ascii plus decimal NCR. In this way we transform Unicode file into an ASCII file in which the Unicode characters are represented using numeric codes &#dddd;. Any of the two files generated can be read by Pajek and proper Unicode characters will appear in Draw, Report and other windows provided that you select proper proportional and monospaced font in Options/Select Font in Pajek Main menu. We suggest to use Arial Unicode MS or Lucida Sans Unicode (one of them usually comes with Windows) for proportional font and GNU Unifont for monospaced font. Courier New can be used too. When saving networks they are stored as ASCII files by default (nonascii characters are stored as &#dddd;) unless you check Options/Read-Write/Save Files as Unicode UTF8 with BOM. The file is used in Pajek 2.00 (or higher) to produce picture. The picture is exported in the SVG format. ανδρει ανδρει Vlado Vlado 焹 焓 燝 牕 犫 獨 焹 焓 燝 牕 犫 獨 गङथफञऑक זלגהשעצ Андреј كصتبخيري זלגהשעצ كصتبخيري गङथफञऑक Андреј Pajek Pajek The SVG picture can be opened in the free drawing program InkScape. Here we can make some additional enhancements of the picture.

81 Exports to EPS/SVG/X3D/VRML 79 In InkScape we can, using the option File/Save as/pdf or eps, save the picture in the desired format PDF or EPS or some other) with labels containing Unicode characters. Note that the PDF (1.4 or higher) also supports the transparency. The EPS file can be inserted in Word document. A set of example files is available in ZIP. Similar approach can be used also for 3D models in X3D. For their inspection we recommend the use of Instant Player.

82 80 Using Macros in Pajek 6 Using Macros in Pajek 6.1 What is a Macro? A macro enables you to record a sequence of primitive Pajek commands into a file. You can use this file later to execute the saved sequence of commands without selecting one by one. 6.2 How to record a Macro? 1. First load all objects (networks, partitions,...) which will be used by a macro. 2. Select objects which will be used in macro (for example, network which will be first used must be shown in Network ComboBox). 3. Choose Macro/Record and select the name of macro file (default extension is.mcr). 4. Use Pajek as usual to define a sequence of commands. We advise to add additional comments to macro file using Macro/Add Message. When running macro this comments are displayed in Report window and help you to follow the macro execution. 5. At the end select Macro/Record again to stop recording. 6.3 How to execute the Macro? 1. First load object(s) that will be used as input in macro execution. 2. Choose Macro/Play and select the macro file. 6.4 Example The following macro performs topological sort of an acyclic network: NETBEGIN 2 CLUBEGIN 1 PERBEGIN 1 CLSBEGIN 1 HIEBEGIN 1 VECBEGIN 1 NETPARAM 1

83 Using Macros in Pajek 81 Msg Depth Partition C 1 DEP 1 (10) Msg Make Permutation P 1 MPER 1 (10) Msg Reordering network N 2 REOR 1 1 (10) The first seven commands store the current state of ComboBoxes. The acyclic network on which we want to execute topological sort must be at the top of Network ComboBox before starting the macro. 6.5 List of macros available in installation file Macros prepared for genealogies and other acyclic networks Path find shortest chain (undirected path) between the two vertices. Descendants find all vertices later than selected vertex (in a p-graph). Ancestors find all vertices before the selected vertex (in a p-graph). Cognatic3 find vertices three generations before and three generations after the selected vertex (in a p-graph). Cognatic find all reachable vertices before and all after the selected vertex (in a p-graph). Layers, Layers1, Layers2 draw acyclic network in layers (different algorithms). zlayers draw acyclic network in layers in z direction. LongestPatrilineage find the longest patrilineage in Ore-genealogy. The genealogy must be read with the option Ore graph: 1-Male, 2-Female links checked. LongestMatrilineage find the longest matrilineage in Ore-genealogy. The genealogy must be read with the option Ore graph: 1-Male, 2-Female links checked. NumDescendants for each vertex compute the size of output domain (number of descendants in Ore-graph). NumAncestors for each vertex compute the size of input domain (number of ancestors in Ore-graph).

84 82 Using Macros in Pajek Macros prepared for computing derived kinship relations Pajek generates three relations when reading genealogy as Ore graph with the option Ore graph: 1-Male, 2-Female links checked: 1 : is a father of 2 : is a mother of 3 : is a spouse of Other kinship relations can be obtained from these relations by running macros. Gender partition is also result of reading genealogy and is input to the following macros. 4 : is a parent of 5 : is a child of 6 : is a son of 7 : is a daughter of 8 : is a husband of 9 : is a wife of 10 : is a sibling of 11 : is a brother of 12 : is a sister of 13 : is an uncle of 14 : is an aunt of 15 : is a semisibling of Using macro add all relations you can add all above relations at once. As a test/example file you can use * /pajek/data/family.ged 6.6 Repeating last command Macro submenu enables to run the last command executed by Pajek several times applied to different successive objects, as well. Example: after loading 100 networks in Pajek, execute degree partition on the first network and run Repeat Last Command by typing 99 to compute degree partition on all other networks. Commands that include several objects can be run as well (e.g. extracting subnetworks according to selected partition(s)). There are 3 different possibilities for extracting from the set of networks (possibilities are selected by fixing appropriate objects): for all networks extracting is determined by the same partition (incrementing applied to network, not applied to partition partition is fixed)

85 Using Macros in Pajek 83 for all networks extracting is determined by another partition (incrementing applied to network and to partition nothing is fixed) for one network several extractions are determined by different partitions (incrementing not applied to network, applied to partition network is fixed) Important: Always first execute the command on the first loaded object(s). Repeating will start with object(s) that have object number(s) one higher than the object(s) on which the command was executed. If the result of the command is also a constant, all constants are stored in a vector. In Pajek the following constants exist: number of vertices, arcs, edges, network densities, centralization indices, diameter, relinking index, correlation and contingency coefficients, number of fragments, main core number, number of components, length of critical path, maximum flow, distance between vertices, number of islands, minimum and maximum value in partition / hierarchy, minimum, maximum, arithmetic mean, median and standard deviation in vector.

86 84 Blockmodeling in Pajek 7 Blockmodeling in Pajek The blockmodeling option is an embedding of the programs for optimization approach to generalized blockmodeling MODEL2 and TwoMODEL from package STRAN STRucture ANalysis [9] into Pajek. The blockmodeling command seeks for the best partition of a given network satisfying given types of blocks generalized blockmodeling [7, 35]. The blockmodel can be built inside Pajek (if User Defined is selected) and/or its description can be stored in MDL file. The impact of errors in each block can be controlled using penalty weights. The option supports also generalized blockmodeling of two-mode networks [34]. The maximum size of a network on which the command can be applied is 250 vertices. But the real limit is time complexity already on 100 vertices optimization can last some hours. The results of the command are stored as partitions. They can be displayed as a picture Draw / Draw-Partition and Layout / Energy / Kamada-Kawai / Free or Layout / Circular / using Partition The result can be displayed also in the matrix form. This requires two steps: Partition / Make Permutation File / Network / Export Matrix to EPS / Using Permutation enter file name; yes 7.1 MDL files The structure of a MDL file is evident from the following example *MODEL Tina *CONSTRAINTS *EOM The first character in each line should be a star * or a blank. The last character in a line should not be a blank.

87 Blockmodeling in Pajek 85 A number in the second line is number of clusters. In the case of two-mode network two numbers should be given in this line number of row-clusters and number of col-clusters. The other data are the same for both type of networks. The following lines have the structure i j penalty t 1 t 2... t k When i, j > 0 the line prescribes that the block (i, j) can be of types t 1, t 2... t k. The types are coded as follows null 7 rfn - row-function 1 com - complete 8 cfn - col-function 2 rdo - row-dominant 9 den - density 3 cdo - col-dominant 10 dnc - do not care 4 reg - regular 11 one - non-null 5 rre - row-regular 12 sym - symmetric 6 cre - col-regular Lines with i = 0 defines the types of of parts of model matrix: j = 0: diagonal (for one-mode networks only); j = 1: upper triangle (for one-mode networks only); j = 2: lower triangle (for one-mode networks only); j = 3: complete matrix. In the case of several lines describing the same block the last prescription prevails. Lines following *CONSTRAINTS define additional constraints in blockmodel. Constraints have the form k penalty i j with the following meaning k = 1: i C j vertex i belongs to cluster C j ; k = 2: i / C j vertex i does not belong to cluster C j ; k = 3: C(i) = C(j) vertices i and j belong to the same cluster; k = 4: C(i) C(j) vertices i and j belong to different clusters; k = 5: i C(j) cluster C j contains at least i vertices;

88 86 Blockmodeling in Pajek k = 6: i C(j) cluster C j contains at most i vertices. Be careful in the case of two mode network for constraints of type k = 5 and k = 6 (constraints on min/max number of vertices in selected cluster): If there are r row-clusters and c col-clusters in blockmodel, then use numbers in the range 1... r to define constraint on row-cluster and numbers in the range r r + c to define constraint on col-cluster. The violations of constraints contribute to criterion function with a term + # of violations penalty The values of penalties have to be in the range 0 to Examples of MDL files Regular blocks *MODEL Regular *EOM Diagonal blocks (clustering) *MODEL Diagonal *EOM Acyclic model (up) *MODEL Hierarchy *EOM Acyclic model with symmetric clusters (down) *MODEL SymHiera 9

89 Blockmodeling in Pajek *EOM Center-Periphery *MODEL Center-Periphery *EOM Regular path *MODEL Regular Path *EOM Regular chain *MODEL Regular Chain

90 88 Blockmodeling in Pajek *EOM mode standard model for Davis.net *MODEL UserDefined *EOM

91 Colors in Pajek 89 8 Colors in Pajek GreenYellow Yellow Goldenrod Dandelion Apricot Peach Melon YellowOrange Orange BurntOrange Bittersweet RedOrange Mahogany Maroon BrickRed Red OrangeRed RubineRed WildStrawberry Salmon CarnationPink Magenta VioletRed Rhodamine Mulberry RedViolet Fuchsia Lavender Thistle Orchid DarkOrchid Purple Plum Violet RoyalPurple BlueViolet Periwinkle CadetBlue CornflowerBlue MidnightBlue NavyBlue RoyalBlue Blue Cerulean Cyan ProcessBlue SkyBlue Turquoise TealBlue Aquamarine BlueGreen Emerald JungleGreen SeaGreen Green ForestGreen PineGreen LimeGreen YellowGreen SpringGreen OliveGreen RawSienna Sepia Brown Tan Gray Black White LightYellow LightCyan LightMagenta LightPurple LightGreen LightOrange Canary LFadedGreen Pink LSkyBlue Gray05 Gray10 Gray15 Gray20 Gray25 Gray30 Gray35 Gray40 Gray45 Gray55 Gray60 Gray65 Gray70 Gray75 Gray80 Gray85 Gray90 Gray95 Figure 19: Colors in Pajek.

92 90 Colors in Pajek 0 - Cyan 1 - Yellow 2 - LimeGreen 3 - Red 4 - Blue 5 - Pink 6 - White 7 - Orange 8 - Purple 9 - CadetBlue 10 - TealBlue 11 - OliveGreen 12 - Gray 13 - Black 14 - Maroon 15 - LightGreen 16 - LightYellow 17 - Magenta 18 - MidnightBlue 19 - Dandelion 20 - WildStrawberry 21 - ForestGreen 22 - Salmon 23 - LSkyBlue 24 - GreenYellow 25 - Lavender 26 - LFadedGreen 27 - LightPurple 28 - CornflowerBlue 29 - LightOrange 30 - Tan 31 - LightCyan 32 - Gray Gray Gray Gray Gray Gray Gray Gray70 Figure 20: Partition colors.

93 Citing Pajek 91 9 Citing Pajek For citing Pajek you may consider: V. Batagelj, A. Mrvar: Pajek Program for Large Network Analysis. Home page: W. de Nooy, A. Mrvar, V. Batagelj: Exploratory Social Network Analysis with Pajek, Structural Analysis in the Social Sciences 27, Cambridge University Press, ISBN: CUP, Amazon. V. Batagelj, A. Mrvar: Pajek Analysis and Visualization of Large Networks. In Jünger, M., Mutzel, P. (Eds.): Graph Drawing Software. Springer (series Mathematics and Visualization), Berlin ISBN Springer, Amazon, preprint V. Batagelj, A. Mrvar: Pajek Program for Large Network Analysis. Connections, 21(1998)2, preprint Figure 21: Gartenkreuzspinne / Araneus diadematus; Photo: Stefan Ernst

94 92 References References [1] Adobe SVG viewer. [2] Ahmed, A., Batagelj, V., Fu, X., Hong, S.-H., Merrick, D., Mrvar, A. (2007): Visualisation and analysis of the Internet movie database. Asia-Pacific Symposium on Visualisation 2007 (IEEE Cat. No. 07EX1615), [3] Albert R., Barabasi A.L.: Topology of evolving networks: local events and universality. [4] Batagelj V.: Papers on network analysis. [5] Batagelj V.: Workshop on Network Analysis, Sydney, Australia: 14th to 17th June 2005; at Nicta (National ICT Australia). [6] Batagelj V.: Some new procedures in Pajek. Dagstuhl seminar 05361, Dagstuhl, Germany, Sept 5-9, [7] Batagelj, V. (1997) Notes on blockmodeling. Social Networks 19, [8] Batagelj V.: Efficient Algorithms for Citation Network Analysis. [9] Batagelj V.: MODEL 2. [10] Batagelj V. (2009): Social Network Analysis, Large-Scale. R.A. Meyers, ed., Encyclopedia of Complexity and Systems Science, Springer 2009: [11] Batagelj V. (2009): Complex Networks, Visualization of. R.A. Meyers, ed., Encyclopedia of Complexity and Systems Science, Springer 2009: [12] Batagelj V., Brandes U. (2005): Efficient Generation of Large Random Networks. Physical Review E 71, , 1-5. [13] Batagelj, V., Ferligoj, A., and Doreian, P. (1992), Direct and Indirect Methods for Structural Equivalence, Social Networks, 14,

95 References 93 [14] Batagelj, V., Doreian, P., and Ferligoj, A. (1992) An Optimizational Approach to Regular Equivalence. Social Networks 14, [15] Batagelj, V, Ferligoj, A, Doreian, P (2007): Indirect blockmodeling of 3-way networks. Selected contributions in data analysis and classification, Springer, Berlin, [16] Batagelj, V., Kejžar, N., Korenjak-Černe, S. (2008): Analysis of the Customers? Choice Networks: An Application on Amazon Books and CDs Data. Metodološki zvezki/advances in Methodology and Statistics 4 (2): [17] Batagelj V., Mrvar A.: Pajek. [18] Batagelj V., Mrvar A. (2000) Some Analyses of Erdős Collaboration Graph. Social Networks, 22, [19] Batagelj V., Mrvar A. (2001) A Subquadratic Triad Census Algorithm for Large Sparse Networks with Small Maximum Degree. Social Networks, 23, [20] Batagelj V., Mrvar A. (2008) Analysis of Kinship Relations With Pajek. Social Science Computer Review 26(2), , [21] Batagelj V., Mrvar A., Zaveršnik M. (1999) Partitioning Approach to Visualization of Large Graphs. In: Kratochvil J. (Ed.) GD 99, Štiřin Castle, Czech Republic. LNCS Springer-Verlag, [22] Batagelj V., Mrvar A., Zaveršnik M. (2002) Network analysis of texts. Language Technologies, Ljubljana, p [23] Batagelj V., Zaveršnik M. (2011): Fast algorithms for determining (generalized) core groups in social networks. Advances in Data Analysis and Classification. Springer [24] Batagelj, V. and Zaveršnik, M. (2007): Short Cycles Connectivity. Discrete Math 307 (3-5): [25] Bollobas B.: Random Walks on Graphs, [26] Broder A. etal. (2000): Graph structure in the web.

96 94 References [27] Burt R.S. (1992): Structural Holes. The Social Structure of Competition. Cambridge MA: Harvard University Press. [28] Butts, C.T. (2002) sna: Tools for Social Network Analysis. [29] Cortona Player (2006) [30] Cosmo Player (2002) [31] de Nooy W., Mrvar A., Batagelj V. (2002) Exploratory Social Network Analysis With Pajek. Structural Analysis in the Social Sciences 27, Cambridge University Press, ISBN: CUP, Amazon, Tokyo Denki University Press [32] Doreian P., Mrvar A. (1996) A Partitioning Approach to Structural Balance. Social Networks, [33] Doreian, P., Batagelj, V., Ferligoj, A. (2000) Symmetric-acyclic decompositions of networks. J. classif., 17(1), [34] Doreian P., Batagelj V., Ferligoj A. (2004) Generalized blockmodeling of two-mode network data. Social Networks 26, [35] Doreian P., Batagelj V., Ferligoj A.: Generalized Blockmodeling, Structural Analysis in the Social Sciences 25, Cambridge University Press, ISBN: CUP, Amazon. [36] Dremelj P., Mrvar A., Batagelj V. (2002) Analiza rodoslova dubrovačkog vlasteoskog kruga pomoću programa Pajek. Anali Dubrovnik XL, HAZU, Zagreb, Dubrovnik, (in Croat). [37] GEDCOM pmcbride/gedcom/55gctoc.htm [38] Ghostscript, Ghostview and GSview. ghost/ [39] Gibbons A. (1985) Algorithmic Graph Theory. Cambridge University Press. [40] Grossman J. (2002) The Erdős Number Project. grossman/erdoshp.html [41] Hall, B.H., Jaffe, A.B. and Tratjenberg M.: The NBER U.S. Patent Citations Data File. NBER Working Paper 8498 (2001).

97 References 95 [42] Hamberger, K., Houseman, M., Daillant, I., White, D.R., Barry, L. (2004) Matrimonial Ring Structures. Math. & Sci. hum. / Mathematics and Social Sciences, 42(4), [43] Hummon, N.P., Doreian, P. (1989) Connectivity in a citation network: The development of DNA theory. Social Networks, 11, [44] Jones B. (2002). Computational geometry database. jeffe/compgeom/biblios.html [45] Kejžar, N., Korenjak Černe, S., Batagelj, V. (2010) Network Analysis of Works on Clustering and Classification from Web of Science. Classification as a Tool for Research. Hermann Locarek-Junge, Claus Weihs eds. Proceedings of IFCS Studies in Classification, Data Analysis, and Knowledge Organization, Part 3, , Springer, Berlin, Springer, preprint [46] Kejžar, N., Nikoloski, Z., Batagelj, V. (2008): Probabilistic Inductive Classes of Graphs. Journal of Mathematical Sociology 32: [47] Kleinberg J. (1998) Authoritative sources in a hyperlinked environment. In Proc 9th ACMSIAM Symposium on Discrete Algorithms, p [48] Knuth, D. E. (1993) The Stanford GraphBase. Stanford University, ACM Press, New York. ftp://labrea.stanford.edu/pub/sgb/ [49] LOCKS: CRA Analyses of News Stories on the Terrorist Attack. Arizona State University. [50] McCabe, T. Computer Science Approaches: Visualization Tools and Software Metrics. in Survey Automation. NAP, 2003, p [51] MDL Information Systems, Inc. (2002) [52] James Moody home page (2002) [53] Murtagh, F. (1985) Multidimensional Clustering Algorithms, Compstat lectures, 4, Vienna: Physica-Verlag. [54] Pajek s datasets:

98 96 References [55] D.M. Pennock etal. (2002) Winners dont t take all, PNAS, 99/8, [56] Peterson J. L.: Petri Net Theory and the Modeling of Systems. [57] The R Project for Statistical Computing. [58] Richardson D.C., Richardson J.S. (2002) The Mage Page. [59] Scott, J. (2000) Social Network Analysis: A Handbook, 2nd edition. London: Sage Publications. [60] Seidman S. B. (1983) Network structure and minimum degree, Social Networks, 5, [61] Tarjan, R. E. (1983) Data Structures and Network Algorithms. Society for Industrial and Applied Mathematics Philadelphia, Pennsylvania. [62] UCINET (2002) [63] The United States Patent and Trademark Office. [64] Wasserman S., Faust K. (1994). Social Network Analysis: Methods and Applications. Cambridge University Press, Cambridge. [65] White D.R., Batagelj V., Mrvar A. (1999) Analyzing Large Kinship and Marriage Networks with Pgraph and Pajek. Social Science Computer Review, 17 (3), [66] Wilson, R.J., Watkins, J.J. (1990) Graphs: An Introductory Approach. New York: John Wiley and Sons. [67] W3C SVG page.